Visit https://ebookultra.

com to download the full version and

explore more ebooks

Bioinformatics and Biomedical Engineering

Proceedings of the 9th International Conference
on Bioinformatics and Biomedical Engineering 1st
Edition James Chou

_____ Click the link below to download _____

https://ebookultra.com/download/bioinformatics-and-
biomedical-engineering-proceedings-of-the-9th-
international-conference-on-bioinformatics-and-
biomedical-engineering-1st-edition-james-chou/

Explore and download more ebooks at ebookultra.com

Here are some suggested products you might be interested in.
Click the link to download

Emergent Computation Emphasizing Bioinformatics Biological

and Medical Physics Biomedical Engineering 1st Edition
Matthew Simon
https://ebookultra.com/download/emergent-computation-emphasizing-
bioinformatics-biological-and-medical-physics-biomedical-
engineering-1st-edition-matthew-simon/

ICSTE 2011 Proceedings of the 2011 3rd International

Conference on Software Technology and Engineering 3rd
Edition International Conference On Software Technology
And Engineering
https://ebookultra.com/download/icste-2011-proceedings-of-
the-2011-3rd-international-conference-on-software-technology-and-
engineering-3rd-edition-international-conference-on-software-
technology-and-engineering/

Proceedings of the 21st International Conference on

Industrial Engineering and Engineering Management 2014 1st
Edition Ershi Qi
https://ebookultra.com/download/proceedings-of-the-21st-international-
conference-on-industrial-engineering-and-engineering-
management-2014-1st-edition-ershi-qi/

Introduction to biomedical engineering 3ed. Edition

Enderle J.

https://ebookultra.com/download/introduction-to-biomedical-
engineering-3ed-edition-enderle-j/
Biomedical Engineering Bridging Medicine and Technology
1st Edition W. Mark Saltzman

https://ebookultra.com/download/biomedical-engineering-bridging-
medicine-and-technology-1st-edition-w-mark-saltzman/

Geological engineering proceedings of the 1st

International Conference ICGE 2007 American Society Of
Mechanical Engineers
https://ebookultra.com/download/geological-engineering-proceedings-of-
the-1st-international-conference-icge-2007-american-society-of-
mechanical-engineers/

Practical Biomedical Signal Analysis Using MATLAB Series

in Medical Physics and Biomedical Engineering 1st Edition
Katarzyna J. Blinowska
https://ebookultra.com/download/practical-biomedical-signal-analysis-
using-matlab-series-in-medical-physics-and-biomedical-engineering-1st-
edition-katarzyna-j-blinowska/

Bioinformatics Software Engineering Delivering Effective

Applications 1st Edition Paul Weston

https://ebookultra.com/download/bioinformatics-software-engineering-
delivering-effective-applications-1st-edition-paul-weston/

Basic Transport Phenomena in Biomedical Engineering 3rd

Edition Fournier

https://ebookultra.com/download/basic-transport-phenomena-in-
biomedical-engineering-3rd-edition-fournier/
Bioinformatics and Biomedical Engineering Proceedings
of the 9th International Conference on Bioinformatics and
Biomedical Engineering 1st Edition James Chou Digital
Instant Download
Author(s): James Chou
ISBN(s): 9781315683010, 1315683016
Edition: 1
File Details: PDF, 11.82 MB
Year: 2015
Language: english
 BIOINFORMATICS AND BIOMEDICAL ENGINEERING

ICBBE15-FM.indd i 9/1/2015 4:36:30 PM

 PROCEEDINGS OF THE 9TH INTERNATIONAL CONFERENCE ON BIOINFORMATICS AND
BIOMEDICAL ENGINEERING, SHANGHAI, CHINA, 18–20 SEPTEMBER 2015

Bioinformatics and Biomedical

Engineering

Editors
James J. Chou
Harvard Medical School, USA

Huaibei Zhou
Wuhan University, China

ICBBE15-FM.indd iii 9/1/2015 4:36:31 PM

 CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2016 Taylor & Francis Group, London, UK

Typeset by V Publishing Solutions Pvt Ltd., Chennai, India

Printed and bound in Great Britain by CPI Group (UK) Ltd, Croydon, CR0 4YY

All rights reserved. No part of this publication or the information contained herein may
be reproduced, stored in a retrieval system, or transmitted in any form or by any means,
electronic, mechanical, by photocopying, recording or otherwise, without written prior
permission from the publisher.

Although all care is taken to ensure integrity and the quality of this publication and the
information herein, no responsibility is assumed by the publishers nor the author for any
damage to the property or persons as a result of operation or use of this publication and/or
the information contained herein.

Published by: CRC Press/Balkema

P.O. Box 11320, 2301 EH Leiden, The Netherlands
e-mail: [email protected]
www.crcpress.com – www.taylorandfrancis.com

ISBN: 978-1-138-02784-8 (Hbk)

ISBN: 978-1-315-68301-0 (eBook PDF)

ICBBE15-FM.indd iv 9/1/2015 4:36:31 PM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Table of contents

Preface xi
Organization xiii
Acknowledgements xv
About the editors xvii

Biomacromolecular sequence, structure and function

Relationship between homo sapiens histamine receptors using data mining 3
S.M. Kim, S.R. Kim & T.S. Yoon
Analysis of Ebolavirus and Marburgvirus using data mining 11
J. Jeong, B. Kim & T.S. Yoon
Identifying the missing protein in human proteome by structure and
function prediction 19
Q.W. Dong & K. Wang
Prediction of SARS coronavirus main protease by artificial Neural Network 27
S.P. Jang, S.H. Lee, S.M. Choi, H.S. Choi & T.S. Yoon

Algorithms, models and applications

Aquatic biodiversity conservation zoning study based on the fuzzy
clustering analysis: A case study of Tieling City, China 35
H. Liu, L. Xu & H. Wang
Ecological security assessment of water source based on the GEM-AHP-GMS model 45
Y. Xin, S.S. Zhang, L. Xu & W. He
Uncertainty of both eco-corridor and centrality originated from resistant surface:
A case study of Liaoning Province, China 53
M.Q. Yu, G.B. Song, S.L. Liu & S.S. Zhang
Continuous blood pressure measurement based on carotid artery diameter 61
H.Y. Li, Y.B. Li, S.L. Li & N. Deng
A novel saccade signals detection algorithm for EOG-based Human Activity
Recognition 67
Z. Lv, J.N. Guan, B.Y. Zhou & X.P. Wu
Hopf bifurcation analysis of a predator-prey model 75
D.D. Nie, Z.L. Xiong & W. Wang
Finding sorting traces of reversals in the presence of hurdles 85
B. Tripathi & Amritanjali
Degree distribution of protein-protein interaction networks formed by gene duplication 91
Z. Ma, H. Xu & X.G. Wu

ICBBE15-FM.indd v 9/2/15 9:08:25 PM

 Computational evolutionary biology
The application of COI and ITS genes in the molecular phylogeny
of water beetles 101
C. Guo, D.J. Bian & L.Z. Ji
Genome-wide identification and phylogenetic analysis of the Ethylene-Insensitive3
(EIN3) and EIN3-Like (EILs) gene family in melons 105
Y. Ma, Y. Tu, X.L. Chen & X.G. Liu

Biomechanics
In vitro and in vivo biomechanical research on cervical arthroplasty and fusion 113
Z.H. Liao & W.Q. Liu
Mechanical conditions affect intervertebral disc degeneration concerning
its water retention 121
W. Weina, W.Q. Liu & Z.H. Liao
Viscoelasticity of the intervertebral segments after fusion under continuous
compression 127
B.Q. Pei, Z.Y. Liu, H. Li & Y.Y. Pei

Biosignal/image processing and analysis

Recognition of sequential upper limb movements based on surface
Electromyography (sEMG) signals 135
B.Y. Zhang, Z.T. Zhou, E.W. Yin, J. Jiang & D.W. Hu
Contrastive study on the enhancement of pitch adjustable Electrolarynx speech
based on spectral subtraction technology 143
Y. Li, L. Wang, Y.J. Feng & H.J. Niu
Comparison between neural network or neural network with genetic algorithm
and analysis of EEG signal 151
Y.A. Rahim & M.M. BkrSudan
Feature optimization for pathological voice based on BP neural network 157
Y. Zeng, W.P. Hu & D.D. Liang
A graph-based method for blood vessel segmentation of retinal images 163
J.D. Zhang, W.H. Jiang, C.X. Zhang & Y.J. Cui
A fusion approach for dynamic skin detection 171
L. Chen & Y.H. Liu
Local gradient thresholds computation in 3D images based
on a vision model 179
J.Y. Li, P. Wang & L.S. Wang
Abnormality detection of specific brain structure in MR images based
on multi-atlas and texture descriptor 187
B.Z. Chen, Y. Wang & L.S. Wang
Retinal images change detection based on fusing multi-features differences 193
S. Yin, B.Z. Chen & L.S. Wang
Effective method for extracting the characteristic value of photoplethysmography
signals 199
Y.Z. Shang, S.S. Yong, D.X. Guo, R. Peng & X.A. Wang

vi

ICBBE15-FM.indd vi 9/1/2015 4:36:31 PM

 Biomedical materials and products
TEOS hydrolysis method synthesize nano silica and its biological
toxicity research 207
L.J. Hu, R. Li, B. Liu, X. Han, X. Zheng & B.B. Tong
Solvothermal method preparation of nano Y2O2S:Eu3+ and research
of luminescence property and cytotoxicity 213
C. Zhou, W.J. Yan, J. Jin, X.X. Li, Y. Li & P. Xie
Enzyme assisted extraction of polysaccharides from Flammulina velutipes
and its radicals scavenging activity 219
G.T. Chen, Y.R. Dong, G.H. Qi, Z.P. Yang, H.X. Wang & S.L. Wang
Preparation of CdTe quantum dots in aqueous phase 227
Y. Hao
Optimization of the preparation of nalmefene hydrochloride injection 233
M.H. Duan, Q. Hao, Y.X. Sun, Z. Zhang, Z. Wang, Y. Yang, L.N. Wang,
Y.H. Jiang, C.L. Zhou & J. Pei
Development and validation of a method for detecting contaminants
in nalmefene hydrochloride injections 239
M.H. Duan, Q. Hao, Y.X. Sun, Z. Zhang, Z. Wang, Y. Yang, Y.H. Jiang,
L.N. Wang, C.L. Zhou & J. Pei
A new technology for the preparation of Deproteinized Extract of Calf Blood 245
H.Y. Li, L.N. Chen, G.X. Yuan, G.Y. Xu, Y. Sheng, P.G. Du, L.P. An & T.C. Li
Preparation of mouse amelotin antibody by synthetic peptides 251
J.J. Zhang & Y. Sun
Optimization of protein depletion technology for polysaccharides from Angelica
and Astragalus by orthogonal test design 257
X.Y. Pu, X.L. Ma, L. Liu, J. Ren, X.Y. Li & H.B. Li

Biomedical devices and systems

Detection of C-Reactive Protein based on a Dynabeads-labeled sandwich
immunoassay by using a GMI biosensor 265
Z. Yang, C. Lei & Y. Zhou
Combining eye gaze detection with vision-based object recognition
for a portable Human-Machine Interface 271
C.X. Yuan, J. Jiang, J.S. Tang, Z.T. Zhou & D.W. Hu
The design of a rehabilitation training system with EMG feedback
for stroke patients 277
C.X. Yu, J.Y. Guo, Z.G. Yu, H.J. Niu, Y.B. Fan, Z. Wang,
W.R. Zhao & H.H. Zhao
Image Guided Surgery system for Optic Nerve Decompression Operation 283
Y.J. Wu, C.W. Xiao, C.L. Fang & L.S. Wang
Denitrification under aerobic condition by a reactor packed with corncobs
as carbon source and bio-carrier 289
L. Shao & Y. Ling
Wireless platform for real-time Electrocardiography (ECG) recording
and analysis 295
M. Al-Qahtani, M. Alwahiby, M. Abdelhamid, E.H. Mirza & X. Yang

vii

ICBBE15-FM.indd vii 9/1/2015 4:36:31 PM

 Design of the Invasive Blood Pressure simulator 301
Q.C. Liu & B. Xiao
Needle guide device development for CT system-based biopsy 307
H.K. Yun, G.R. Park, K.C. Choi, T.S. Shin, M.K. Kim & S.K. Joo
A novel target selection approach by incorporating image segmentation
into P300-based Brain-Computer Interfaces 313
P. Du, Y. Yu, E.W. Yin, J. Jiang, Y.D. Liu & D.W. Hu
Rapid identification of Panax ginseng and Panax quinquefolius using SNP-based
probes 319
R.L. Wang, D.J. Gu, C.R. Hou & Q.J. Liu
Study on the range of motion of laparoscopic instruments with a simulator 325
K. Lu, C. Song & C. Wei

Biomedical imaging
The application of photoacoustic tomography in joint tissues 335
X.C. Zhong, X.Y. Jing, S.Q. Jing, N.Z. Zhang & J. Rong
Frequency response mismatch correction in multichannel time interleaved analog
beamformers for ultrasound medical imaging 341
A. Zjajo & R. van Leuken
HIFU based photoacoutic tomography 349
X.C. Zhong, W.Z. Qi, S.Q. Jing, N.Z. Zhang & J. Rong
Analyzer-Based Phase Contrast X-ray Imaging for mouse tissues 355
H. Li, M. Wang, Z. Wang & S.Q. Luo

Biocybernetics and biological effects

Effects of captopril pretreatment on gp130 expression in rats
with acute cardiomyocyte injury 361
Y. Zhou, S.B. Li, Y. Zhang, X.B. Jing & M. Gu
Effects of sleep restriction and circadian rhythm on human vigilance 367
H.Q. Yu, Y. Tian, C.H. Wang, S.G. Chen & J.H. Guo
The impact of environmental factors on the survival status of Rana 375
M.H. Duan, X.H. Li, A.J. Jiang, Y.Y. Pei, D.D. Guan, Y.X. Sun, H. Qiang,
Y. Yue & Y.P. Li
Balance role of dopamine system on exciting and inhibitory involved in motor
function 381
J.L. Cheng, X.W. Wang & L.J. Hou
Pemetrexed and carboplatin chemotherapy combined with whole brain
radiotherapy for non-small cell lung cancer with brain metastasis 387
H.B. Li, Y.F. Yun, X.F. Zhou, J.D. Luo, H. Zhu, L.Y. Zhou & Q. An
Metabolites analysis of functional lactic acid bacteria strain Lactobacillus
paracasei HD1.7 395
R.P. Du, D. Zhao, X.Y. Wang, Q. Wang & J.P. Ge
4-aminopyridine inhibits cell proliferation and affects anti-tumor activities
of cell cycle-specific drugs in human breast cancer cells 403
Q. Ru, X. Tian, J.S. Liu & C.Y. Li

viii

ICBBE15-FM.indd viii 9/1/2015 4:36:31 PM

 Biostatistics and biometry
Epidemiological characteristics of outpatients in a Grade 3, Class A general hospital 411
Y.M. Li, Y.J. Tan, F. Wu, C. Zheng, N.Y. Sun & K.Y. Wang
Methane estimation of food waste in Chinese household and environmental
benefits from an energy perspective 417
S. Ding, G.B. Song & S.S. Zhang
A method of Chemiluminescence Enzyme Immunoassay for Zearalenone 423
K.H. Li, L.X. Zhu, W. Meng & R.R. Liu
Estimation of blood glucose noninvasively using near infrared spectroscopy 429
R. Peng, D.X. Guo, Y.Z. Shang, S.S. Yong & X.A. Wang
Restrictive factors of the measurement accuracy to estimate Blood
Pressure with Pulse wave Transit Time 435
X.M. Chen, Y.B. Li, Y. Zhang & N. Deng

Rehabilitation engineering
The research on motion recognition based on EMG of residual thigh 445
T.Y. Zhang

Author index 451

ix

ICBBE15-FM.indd ix 9/1/2015 4:36:31 PM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Preface

It is our great pleasure to present the proceedings of The 9th International Conference on
Bioinformatics and Biomedical Engineering (iCBBE 2015), held September 18–20, 2015 in
Shanghai, China. We would like to take this opportunity to express our sincere gratitude and
appreciation to all the authors and participants for their support of this conference.
The research on Bioinformatics and Biomedical Engineering has enormous impacts on
science, education, culture and society as well. Actually, the discipline of Bioinformatics and
Biomedical Engineering has become a new focus of life science, mathematical science, com-
puter science and electronic information science. More and more scientists all around the
world are dedicating themselves to this interdisciplinary area, accumulating a lot of interest-
ing results.
We are proud to see that the previous iCBBE conferences were successful in providing an
ideal platform for them to exchange their exciting findings, to stimulate the further develop-
ment of Bioinformatics and Biomedical Engineering, and to enhance its impacts to various
areas of both science and medicine (see, e.g., Medicinal Chemistry, 2015, 11, 218–234). We
believe that the 2015 iCBBE will do even better in this regard.
On behalf of the organizing committee, we would like to take this opportunity to express
our gratitude to the conference’s sponsors: Wuhan University and The Gordon Life Science
Institute.
Our appreciation and gratitude are also extended to all the papers’ reviewers and the Con-
ference Organization Committee members. It is impossible to hold such a grand conference
without their help and support.
The papers collected in the “2015 iCBBE Proceedings” provide the detailed results of some
oral presentations that will be of use to the readership.

Editors
Prof. James J. Chou
Harvard Medical School, USA

Prof. Huaibei Zhou

Wuhan University, China
2015

xi

ICBBE15-FM.indd xi 9/1/2015 4:36:31 PM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Organization

This volume contains the Proceedings of the 9th International Conference on Bioinformat-
ics and Biomedical Engineering (iCBBE 2015)—held September 18–20, 2015 in Shanghai,
China. iCBBE2015 has been organised by Wuhan University and The Gordon Life Science
Institute.

INTERNATIONAL PROGRAMME COMMITTEE

Honorary General Chair

Prof. Kuo-Chen Chou, The Gordon Life Science Institute, USA

General Chair
Prof. James J. Chou, Harvard Medical School, USA

Technical Program Committee

Prof. Shu Q. Liu, Northwestern University, USA
Prof. Fengfeng Zhou, Chinese Academy of Sciences, China
Prof. Ridha Hambli, Orleans University, France
Dr. Fadhl M. Al-Akwaa, University of Science and Technology, Yemen
Dr. Yu Chen, University of Strathclyde, UK
Prof. Musa Hakan Asyali, Antalya International University, Turkey
Prof. Lukasz Kurgan, University of Alberta, Canada
Prof. Huabei Jiang, University of Florida, USA
Prof. Jerzy Tiuryn, University of Warsaw, Poland
Dr. Joseph Chang, Nanyang Technological University, Singapore
Dr. Mengxing Tang, Imperial College London, UK
Dr. Yanmei Tie, Harvard Medical School, USA
Dr. Manuchehr Soleimani, University of Bath, UK
Prof. Jinn-Moon Yang, National Chiao-Tung University, Chinese Taipei
Dr. Humberto González-Díaz, University of the Basque Country, Spain
Dr. Deligianni Despina, University of Patras, Greece
Dr. Shuwei Li, Ambry Genetics, USA
Dr. Suryani Lukman, Khalifa University of Science, Technology and Research, UAE
Prof. Sheng-Xiang Lin, Laval University Medical Center, Canada
Dr. Wei-Zhu Zhong, The Gordon Life Science Institute, USA
Prof. Rajiv Mahendru, BPS Government Medical College for Women, India
Prof. Zodwa Dlamini, University of South Africa, South Africa

xiii

ICBBE15-FM.indd xiii 9/1/2015 4:36:31 PM

 LOCAL ORGANIZING COMMITTEE

Fang Liu, Wuhan University, China

Liang Li, Wuhan University, China
Ruoshan Kong, Wuhan University, China
Xiaoyan Sheng, Wuhan University, China
Yuanyuan Cheng, Wuhan University, China
Yujing Zhang, Wuhan University, China

xiv

ICBBE15-FM.indd xiv 9/1/2015 4:36:31 PM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Acknowledgements

The Organising Committee members wish to express their sincere gratitude for the finan-
cial assistance from the following organisations: Wuhan University, the Gordon Life Science
Institute and the 1000 Think Tank.
The technical assistance of all paper peer reviewers and the publisher CRC Press/Balkema
is gratefully acknowledged. We are also thankful to the International Programme Commit-
tee as well as the members of the Local Organising Committee. Finally, the editors want to
acknowledge all peer reviewers for their great efforts and contributions to us.

Editors
Prof. James J. Chou
Harvard Medical School, USA

Prof. Huaibei Zhou

Wuhan University, China

xv

ICBBE15-FM.indd xv 9/1/2015 4:36:31 PM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

About the editors

Prof. James J. Chou, Harvard Medical School, USA

James J. Chou is a Professor in Biological Chemistry and Molecular Pharmacology at Har-
vard Medical School. He received his BS in Physics from University of Michigan at Ann
Arbor; Ph.D. in Biophysics from Harvard University; and postdoctoral training from NIH.
Since 2002, he joined the faculty at Harvard Medical School. Professor Chou has received
many prestigious awards such as the Pew Scholar Award in Biomedical Sciences, the Smith
Family Award, and the Genzyme Outstanding Achievement in Biomedical Science Award.

Prof. Huaibei Zhou, Wuhan University, China

Prof. Huaibei Zhou received his B.S. in Radio Wave Propagation and Antenna Design in
Wuhan University, China in 1984; Ph.D. in Computational Physics from the University of
Maryland at College Park and Post-Doc in Biotechnology in the National Institute for Stand-
ard and Technology in 1994; and MBA in Engineering Management from George Washington
University in 2001. Prof. Huaibei Zhou discovered the chaotic motion of protein molecules
by simulating the dynamics of protein molecules in water (the simulation is programmed in
C and run in Unix environment), the result has been published as the first paper in this field;
more than 30 scientists around the world have cited this paper in their publications.

xvii

ICBBE15-FM.indd xvii 9/1/2015 4:36:31 PM

 Biomacromolecular sequence, structure and function

ICBBE15_Book.indb 1 8/25/2015 7:20:20 AM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Relationship between homo sapiens histamine receptors using

data mining

S.M. Kim, S.R. Kim & T.S. Yoon

Department of Natural Science, Hankuk Academy of Foreign Studies (HAFS), Yongin, South Korea

ABSTRACT: In this study, we tried to find rules between mRNA sequences of Homo
sapiens histamine receptors H1, H2, H3 and H4 with data mining algorithms, namely Apriori
and Decision tree. In the Apriori algorithm, we split sequences into 5, 7 and 9 windows. The
results showed a strong relationship between the H1 and H4 receptors, and also between the
H2 and H3 receptors. The receptors were divided into two groups according to their compo-
nents. Additionally, we would leave relevancy between the H2 and H3 receptors for another
study with a different data mining algorithm. In the case of the H1 and H4 receptors, we
found that amino acid “F(phenylalanine)” would be a standard to classify the H1 and H4
receptors. We suggest that H4 could be a mutated form of H1. To support our hypothesis, we
conducted an additional experiment with the Decision Tree algorithm, focusing on the exist-
ence of amino acid “F”. The data showed the difference between the H1 and H4 receptors.
In conclusion, the H4 and H1 receptors are related to each other by mutation.

1 INTRODUCTION

Immune system is the system of the animal body that acts as a protection system against
the pathogen from the outer environment. However, there are several diseases caused by
disruptions in immune system functions, such as allergies. Allergies are a result of exagger-
ated and hypersensitive responses to certain antigens called allergens. Antigens of the IgE
class involves in allergic reaction. IgE antibodies are attached to mast cells in connective
tissues by their base. When antigens enter the body, they attach to antigen-binding sites of
IgE antibodies. This connects other antigens near IgE antibodies, causing these antigens to
band together. Mast cells secrete histamines or other inflammation induction substances
from granules. This process is called degranulation. Histamine expands the blood vessel
and increases the permeability of capillaries, which causes typical allergy symptoms such
as sneezing, runny nose, watery eyes and smooth muscle contractions which may lead to
breathing difficulty. Antihistamines blocks histamines from combining to receptor in order
to diminish allergy symptoms (Jane. 2011). Histamine is a type of amine that is produced
and secreted in the animal immune system. It is used in local immune response in order to
cause inflammation. Histamine, which is secreted from mast cells, would bind to its receptor
(histamine receptor). There are four kinds of histamine receptor known in the human body,
which are named H1, H2, H3 and H4. Previously, it was known that histamine receptor 1
(H1) is highly involved in allergic reaction; however, recent studies have shown that hista-
mine receptor 4 (H4) is also involved. It is supposed that there may be some kind of relation-
ships between these two receptors (Thrumond. 2008) (Fung-Leung. 2004). In this study, we
compare the amino acid strand of four types of histamine receptor H1, H2, H3 and H4 by
using an Apriori algorithm and a Decision Tree algorithm in order to see relatedness and
isoforms between the receptors.

ICBBE15_Book.indb 3 8/25/2015 7:20:20 AM

 2 MATERIALS AND METHODS

2.1 Materials
For the experiment, we collected the mRNA sequence of Homo Sapiens histamine receptors
H1 (HRH1)*, H2 (HRH2)**, H3 (HRH3)*** and H4 (HRH4)**** including its transcript
variants. The mRNA sequences used can be found in the NCBI database.

2.2 Methods
To process the data, we used two kinds of algorithms: Apriori and Decision Tree (Lee. 2014)
(Lim. 2014) (Go. 2014).

2.2.1 Apriori algorithm

Apriori algorithm is usually used in data processing and in the field of bioinformatics to find
rules among the continuative data (Kim. 2014). It shows the volume of common parts exist-
ing in the data. By using the Apriori algorithm, we can clearly compare objects with high
frequency in certain data with those found in other data. In this study, 5, 7, and 9 windows
were used to search for amino acid showing high frequency in each data. Also, broken data,
which are rarely found during data processing, were not used in the experiment. This is the
sample of the result of the 5 window Apriori algorithm experiment (Kwon. 2014).

Best rules found:

1. amino5=L 42
2. amino3=L 40
3. amino4=L 35
4. amino1=L 33
5. amino4=S 31
In the example, it showed amino5=L 42 as the first rule. This means that amino acid
L (Leucine, Leu) is the most frequent amino acid in the 5 window as it showed 42 times
repeatedly. Also, statistical compilations focused on the frequency and number of amino acid
appearance instead of its location. According to these data, we constructed a graph in order
to see the data intuitively.

*“Homo sapiens histamine receptor H1 (HRH1), transcript variant 1, mRNA”, NCBI Reference
Sequence: NM_001098213.1 (4,578 bp linear mRNA).
“Homo sapiens histamine receptor H1 (HRH1), transcript variant 2, mRNA”, NCBI Reference
Sequence: NM_001098212.1 (4,298 bp linear mRNA).
“Homo sapiens histamine receptor H1 (HRH1), transcript variant 3, mRNA”, NCBI Reference
Sequence: NM_001098211.1 (4,348 bp linear mRNA).
“Homo sapiens histamine receptor H1 (HRH1), transcript variant 4, mRNA”, NCBI Reference
Sequence: NM_000861.3 (4,427 bp linear mRNA).
**“Homo sapiens histamine receptor H2 (HRH2), transcript variant 1, mRNA”, NCBI Reference
Sequence: NM_001131055.1 (2,624 bp linear mRNA).
“Homo sapiens histamine receptor H2 (HRH2), transcript variant 2, mRNA”, NCBI Reference
Sequence: NM_022304.2 (3,095 bp linear mRNA).
***“Homo sapiens histamine receptor H3 (HRH3), mRNA”, NCBI Reference Sequence: NM_007232.2
(2,680 bp linear mRNA).
****“Homo sapiens histamine receptor H4 (HRH4), transcript variant 1, mRNA”, NCBI Reference
Sequence: NM_021624.3 (3,686 bp linear mRNA).
“Homo sapiens histamine receptor H4 (HRH4), transcript variant 2, mRNA”, NCBI Reference
Sequence: NM_001143828.1 (3,422 bp linear mRNA).
“Homo sapiens histamine receptor H4 (HRH4), transcript variant 3, mRNA”, NCBI Reference
Sequence: NM_001160166.1 (3,522 bp linear mRNA).

ICBBE15_Book.indb 4 8/25/2015 7:20:20 AM

 2.2.2 Decision tree algorithm
While the Apriori algorithm is used to extract the frequency of various amino acids, Deci-
sion Tree is an effective alternative method to show the difference between the data clearly
or definitely (Lim 2014). With the Decision Tree algorithm, isoform data were gathered from
those sequences that were not able to classify with the Apriori algorithm. We determined
that the Decision tree Algorithm is the most appropriate data mining algorithm and well-
ordered lists, which can perform the function of classifying data and finding classes that
represent ability. We compared the histamine receptors H1 and H4, which were expected to
give clear data. We divided into 3 window sizes (5, 7 and 9) and split the data up to 7 classes.
Class 1 to 4 each responds to the Homo sapiens histamine receptor H1 (HRH1) transcript
variant 1 through 4, and class 5 to 7 each responds to the Homo sapiens histamine receptor
H4 (HRH4) transcript variant 1 through 3. In this study, we used a 10-fold cross-validation
method and rule-based classifiers. To retain accuracy, sets of the rules with a frequency over
0.800 in each class were only used to obtain the conclusion.

3 DISCUSSION

3.1 Apriori algorithm

From the result of the Apriori algorithm, we obtained two significant results. The hista-
mine receptors may be divided into two groups by their amino acid, which would put the
H1 and H4 receptors in the same group and the H2 and H3 receptors in the other group.
From Table 1, for each histamine receptor, high frequencies of L (Leucine, Leu) and S
(Serine, Ser) amino acid were found. As these amino acids exist commonly in all receptors,
we defined these two amino acids as a “Basic Group”. In contrast, there were amino acids
that were only found in a certain type of histamine receptors such as A (Alanine, Ala), F
(Phenylalanine, Phe) and G (Glycine, Gly). A and G amino acids were found frequently in
the H2 and H3 receptors; however, F amino acid was only found in the H4 receptor. We
defined these amino acids as an “Additional Group”. We define that “Basic Group” amino
acids would carry out a common function of histamine receptor, while “Additional Group”
amino acids assigns a distinct function to the receptor. According to this definition, we sug-
gest that the H2 and H3 receptors take a similar position in the immune system such as local
inflammation, which differs from the H1 and H4 receptors that takes position in allergy
reaction. As the H1 receptor only has peaks of “Basic Group” amino acids, we would name
the receptor as a type 0 receptor. Also, the H4 receptor is named as type 0` receptor since it
has only one additional amino acid (F) compared with H1 that does not overlap with any
other receptors. As this receptor differs only in F amino acid, we believe this single amino

Figure 1. Graph of HRH1_5,7,9 window.

ICBBE15_Book.indb 5 8/25/2015 7:20:20 AM

 Figure 2. Graph of HRH2_5,7,9 window.

Figure 3. Graph of HRH3_5,7,9 window.

Figure 4. Graph of HRH4_5,7,9 window.

ICBBE15_Book.indb 6 8/25/2015 7:20:20 AM

 Table 1. Rule extraction under 5 window.
Histamine receptor Rule Frequency

1-1 pos2 = D pos3 = K 0.800

pos3 = K pos5 = P 0.800
pos1 = C pos5 = Y 0.800
1-2 pos2 = R pos5 = D 0.800
pos1 = Q pos5 = P 0.800
pos1 = H pos5 = A 0.857
1-3 pos2 = G pos3 = K 0.800
pos3 = E pos4 = N 0.800
pos2 = Q pos5 = V 0.833
pos2 = S pos4 = P pos5 = L 0.800
1-4 pos2 = S pos4 = S pos5 = C 0.800
pos1 = S pos2 = L pos3 = L 0.800
pos1 = S pos3 = S pos5 = W 0.800
pos3 = C pos4 = T 0.800
pos3 = M pos5 = A 0.800
4-1 Not extracted
4-2 pos1 = P pos5 = T 0.800
pos2 = K pos4 = F pos5 = L 0.833
4-3 pos2 = T pos5 = P 0.800
pos1 = H pos4 = V 0.800

Table 2. Rule extraction under 7 window.

Histamine receptor Rule Frequency

1-1 pos1 = K pos2 = A 0.800

pos2 = M pos4 = E 0.800
pos1 = S pos2 = K 0.800
1-2 pos1 = W pos2 = A 0.800
pos1 = C pos2 = T 0.800
pos2 = W pos6 = R 0.800
pos2 = L pos6 = Q 0.800
1-3 Not extracted
1-4 pos1 = M pos2 = L 0.800
pos1 = I pos2 = T 0.800
pos2 = C pos3 = T 0.800
pos2 = G pos5 = Q 0.800
4-1 pos2 = I pos7 = I 0.800
pos2 = S pos5 = Q pos7 = L 0.800
pos2 = N pos4 = S 0.800
4-2 pos3 = F pos4 = F 0.800
4-3 Not extracted

acid makes a small difference between H1 and H4 amino acids, while they still have a more
common part. This explains why the H1 and H4 receptors share a common position in the
immune system even though they are named differently. Also, the difference between the H2
and H3 receptors would occur by the amount of additional amino acids P and A. Although
the H2 and H3 receptors share their additional amino acids A, P and G, the H2 receptor has
7

ICBBE15_Book.indb 7 8/25/2015 7:20:21 AM

 Table 3. Rule extraction under 9 window.
Histamine receptor Rule Frequency

1-1 pos2 = A pos9 = I 0.800

pos5 = R pos8 = S 0.800
pos2 = L pos5 = P 0.857
1-2 pos5 = L pos7 = G pos9 = G 0.800
pos1 = S pos3 = T 0.857
pos1 = W pos3 = T 0.800
pos1 = S pos5 = C 0.857
1-3 pos7 = V pos9 = G 0.800
pos8 = I pos9 = W 0.800
pos4 = T pos6 = L 0.800
pos2 = E pos8 = E 0.800
pos1 = P pos8 = H 0.800
pos3 = L pos9 = Q 0.800
pos8 = A pos9 = P 0.800
pos2 = E pos5 = T 0.800
1-4 pos7 = G pos9 = A 0.800
pos5 = T pos9 = T 0.800
pos4 = E pos9 = Y 0.800
pos3 = C pos5 = C 0.800
pos4 = N pos8 = R 0.800
4-1 pos5 = L pos7=L pos9 = L 0.800
4-2 pos6 = V pos8 = N 0.800
4-3 pos6 = T pos9 = N 0.817
pos2 = R pos8 = L 0.857

shown a relatively high peak on P amino acid, while H3 amino acid had a peak on A amino
acid. As this difference cannot be defined through the Apriori algorithm, we believe that we
have reached the limit of the Apriori algorithm. So, with other data mining algorithm, this
difference may be defined.

3.2 Decision tree algorithm

Decision Tree algorithm was used in order to find an additional relationship between the
H1 and H4 receptors. The purpose of using the Decision Tree algorithm is to find another
rule that did not show the Apriori Algorithm. As stated before, data with a frequency over
0.800 were only used. This is because there were too many data with a frequency below
0.800 for the list or sequence analysis and the finding rule. The most important point of this
experiment is to find a new rule that we were not able to find using the Apriori algorithm
and about F amino acid that belong to the “Additional Group” that we defined in the first
experiment. In this way, we obtained an interesting data. It is that F amino acid was found at
high frequency in the result of experiment by window 5 and window 7 about class 6, which
is an experiment of the H4 receptor. This is an evidence about our suggestion to define the
H1 and H4 receptors as type 0 and type 0` receptors by finding similarity with those in the
Apriori experiment.

ICBBE15_Book.indb 8 8/25/2015 7:20:21 AM

 4 CONCLUSION

Through two experiments using the Apriori algorithm and the Decision Tree algorithm, we
found some important facts. Based on these facts, we made some deduction on the relationship
between the H2 and H3 receptors and also on the relationship between the H1 and H4 recep-
tors. The relationship between the H2 and H3 receptor is the first one that we have found.
In the experiment with the Apriori algorithm, we found a strong correlation between the H2
and H3 receptors. Both receptors are composed of the same types and frequency of amino
acid, which leaves the only difference in the small frequency change of “Additional” amino
acid. In other words, this means that another type of data mining algorithm and experiment
is required in order to specify the difference. We would leave this part as possibility for the
next study and experiment.
The second one is about the H1 and H4 receptors, which is the most important point in
our experiment. In the experiment using the Apriori algorithm, we found a large correlation
and a small difference between the H1 and H4 receptors, which is whether they did or did not
have F amino acid. We decided that an additional experiment is needed in order to find more
information. Decision Tree algorithm was chosen for the second experiment. The results did
not differ much from the experiment using the Apriori algorithm, which became a big ground
for our hypothesis, which is evolutionary variation. This supports our suggestion to be valid.
In conclusion, we have analyzed the mRNA sequence of Homo sapiens histamine receptors,
which is related to the cause of allergy. The results showed much relevancy between the H1,
H2, H3 and H4 receptors. Also, we defined some of receptors to verify our hypothesis. As
there are results consistent with the experimental result data, we found that our study was
certainly meaningful. Our future task of this study is to develop an integration system for a
better environment in studies in the field of bioinformatics.

REFERENCES

Fung-Leung W.P. 2004. Histamine H4 receptor antagonists: the new antihistamines? Current Opinion
in Investigational Drugs.
Go E.B 2014. Analysis of Ebolavirus. International Journal of Machine Learning and Computing
(IJMLC).
Jane B.R. 2011. BIOLOGY, 9th Edition. California, CA: Pearson Education Inc.
Kim D.Y. 2014. Comparison of Hemagglutinin and Neruaminidase of Influenza A Virus Subtype H1N1,
H5N1, H5N2, and H7N9 using Apriori Algorithm. Lecture Note Computer Science (LNCS).
Kwon J.W 2014. Comparison of HTLV and STLV. APCBEE Procedia.
Lee J.H. 2014. Analysis of Malaria Inducing P. Falciparum P. Ovale, and P. Vivax. APCBEE Procedia
Lim S.J. 2014. Analyzing Patterns of Various Avian Influenza Virus by Decision Tree. International
Journal of Computer and Electrical Engineering (IJCEE).
Lim S.J. 2014. rRNA of Alphaproteobacteria Rickettsiales and mtDNA Pattern Analyzing with Apriori &
SVM. Lecture Note Computer Science (LNCS).
Thrumond R.L. 2008. The role of histamine H1 and H4 receptors in allergic inflammation: the search
for new antihistamines. Nat Rev Drug Discov.

ICBBE15_Book.indb 9 8/25/2015 7:20:21 AM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Analysis of Ebolavirus and Marburgvirus using data mining

J. Jeong, B. Kim & T.S. Yoon

Department of International, Hankuk Academy of Foreign Studies, Yongin-si, South Korea

ABSTRACT: Ebolaviruses and marburgviruses, which cause viral haemorrhagic fevers in

humans and often prove to be fatal, are part of the Filviridae family. The virus family was
first identified in 1967 from green monkeys by laboratory workers in Germany and Yugo-
slavia and since then, there have been several outbreaks and large epidemics related to it. In
this paper, we compared the protein sequences encoded in the RNA genomes of ebolaviruses
and marburgviruses particularly focusing on VP24, VP35, and GP. In addition, by examining
the position and frequency of each amino acid, we analyzed the similarities between the two
members of the filviridae family.

1 INTRODUCTION

Filvirdae is a virus family of RNA viruses belonging to the order Mononegavirales. Ascribed
to causing hemorrhagic fever in primates including both human and non-humans, members
of the family (called filoviruses or filovirids) are identified as hazardous by organizations.
The family Filovirade contains three virus genera: Cuevavirus, Ebolavirus, Marburgvirus.
Ebolavirus can be further divided into five species: Bundibugyo ebolavirus (BEBOV), Reston
ebolavirus (REBOV), Sudan ebolavirus (SEBOV), Tai Forest ebolavirus (CIEBOV/TAFV),
and Zaire ebolavirus (REBOV) (Henzy, 2014). Ebolavirus and marburgvirus are thought
to be zoonotic, passed from animals to humans. Fruit bats of the Pteropodidae family are
generally considered to be the host of both viruses; however, apes and chimpanzees are also
regarded as possible hosts. More specifically, scientists infer that rosettus aegypti, which are
fruit bats inside the Pteropodidae family, are the natural hosts of Marburgvirus, transmitting
the disease to people and causing it to spread to other humans. The viruses both cause symp-
toms such as malaise, muscle pain, sudden fever, headache, etc., and at the moment, there is
no standardized cure of filovirus diseases (WHO, 2014). The name Filviridae comes from the
Latin term filum, meaning “threadlike” which accurately describes the slender structure of
filovirions. Filoviruses contain linear, non-segmented, single-stranded and antisense (often
called negative-sense) RNA genomes, ∼19-kb long in length. Proteins NP, VP30, VP35, and
L form the nucleocapsid; proteins VP24 and VP24 form the viral matrix, and protein GP
forms the surface of the particle. In total, filoviruses contain seven proteins that function in
its own distinct way.

2 MATERIALS AND METHODS

As mentioned above, Filovirade can be classified into three virus genera: Cuevavirus,
Ebolavirus, Marburgvirus. This research focuses on the latter two, which are the more com-
monly known of the three. By using decision tree and apriori algorithm, we tried to obtain
the similarities and the differences between the protein sequences of the Ebolavirus and Mar-
burgvirus by analyzing proteins VP24, VP35, and GP from the RNA genome.

11

ICBBE15_Book.indb 11 8/25/2015 7:20:21 AM

 2.1 VP24, VP35
VP24 and VP35 target Interferon-Stimulated Genes (ISG) which are proteins produced by
interferon, used to combat viruses. Though they attack from different angles, VP24 and VP35
both threaten the host cell’s immune system. VP24 block the immune system from transcrip-
tion, discouraging signaling molecules from reaching the cell nucleus (Xu et al., 2014), and
VP35 blocks the path of Type I IFN, keeping them from being produced (Prins et al., 2009).

2.2 GP (Glycoprotein)
The Glycoprotein (GP) is in charge of letting in the virus into target host cells. Located on
the surface of the particle, GP mediates the entry.

2.3 Decision tree

A decision tree is a model used to graphically outline the possible consequences of decisions.
It consists of three nodes: decision nodes, chance nodes, and end nodes. Drawn from left to
right, the branches show the possible actions and the path from root to leaf shows the clas-
sification rules. Since a decision tree is a method of displaying algorithm, it is possible to
analyze decisions, calculate the probability, and figure out the most optimal strategy; there-
fore, it is often used in operations research (Rokach, Maimon, 2010). Traditionally, it was
drawn manually, but nowadays, graphics program and software are utilized to create it. In
this paper, the decision tree was composed of window 5, 7, 9 and experimented with groups
VP24, VP35, and GP. We were unable to find the rules of the decision tree when the subjects
were too similar to one another.

2.4 Apriori algorithm

Apriori algorithm is the simplest set class theory used to find the frequency of the elements.
and the association rules from data. It uses a “bottom up” approach to test candidates against
the data, ending when there are no subset extensions left (Shi, 2011). Like the decision tree,
the experiment was carried out with windows 5, 7, and 9. The results show the position where
certain types of amino acids most frequently appeared.

3 EXPERIMENT

3.1 Decision tree

5 window shows that there is no particular relationship between the GP of Ebola virus and
Marburg virus under window 5. Also, according to Table 1, like the Ebola virus, in Marburg
virus, position 7 and position 3 occurs most frequently. This implies position 7 and position
3 play important roles in both viruses. Table 2 clearly shows that rules were only extracted in
position 8. However, in the case of Ebola virus, pos8 = E was the most frequent, while in the
case of Marburg virus, pos8 = S was the most frequent.
According to Table 3, the result of rule extraction of VP24 under 5 window, rules are
only extracted at position 4. When looking at the graph as a whole, the frequency of amino
acid extracted at position 4 of Ebola virus is higher than that of Marburg virus. No other

Table 1. Rule extraction of GP under 7 window.

Virus Rule

Ebola Position 7 = S, Position 3 = H

Marburg Position 7 = T, Position 1 = I,
Position 3 = A, and Position 7 = I

12

ICBBE15_Book.indb 12 8/25/2015 7:20:21 AM

 Table 2. Rule extraction of GP under 9 window.

Virus Rule

Ebola Position 8 = Q, Position 8 = E, Position 8 = M

Marburg Position 8 = Y, Position 8 = S, Position 8 = F

Table 3. Rule extraction of VP24 under 5 window.

Virus Rule

Ebola Position 4 = A, Position 4 = P, Position 4 = R

Marburg Position 4 = I, Position 4 = N

Table 4. Rule extraction of VP35 under 5 window.

Virus Rule

Ebola Position 1 = D, Position 5 = G

Marburg Position 3 = H, Position 1 = D, Position 2 = K

Table 5. Rule extraction of VP35 under 7 window.

Virus Rule

Ebola Position 1 = Y, Position 1 = N

Marburg Position 1 = A, Position 1 = E

relationships are proven. No rules can be extracted from VP24 of both viruses under
windows 7 and 9.
Table 4 conveys that there are similarities in pos1 = D between VP35 of Ebola virus and
Marburg virus. Although their frequency is different from 0.833 to 0.800, sharing of the
amino acid in this position might deeply contribute to the similarity of two viruses. According
to Table 5, rules are extracted only from the position 1 under 7 window. So, we assume that
position 1 will be the crucial factor which makes Ebola and Marburg virus different from
each other. Lastly, no rules can be extracted under 9 window.

3.2 Apriori algorithm

The results driven from Apriori algorithm are shown below:

Position 4 = T 19, Position 5 = T 18, Position 2 = D 16, Position 1 = T 14

Figure 1 is the result shown by analyzing GP (Glycoprotein) of Ebolavirus under 5 windows.

Analyzing other proteins of either virus had similar results with this. We chose the most fre-
quent rules of each protein of each virus. As we want to show similar or analogous patterns
in each protein of both viruses, we organized the types of amino acid by the frequency of
them.
Figure 2, Figure 3, and Figure 4 are the analysis of GP of Ebolavirus and Marburgvirus
under windows 5, 7, and 9. 5-window: Comparison between Ebolavirus amino4 Thiamine
and Marburgvirus amino1 Thiamine. 7-window: Ebolavirus amino6 Thiamine and Marburg-
virus amino4 Thiamine. 9-window: Ebolavirus amino6 Thiamine and Marburgvirus amino5
Thiamine. In all 3 windows, Marburgvirus has higher Thiamine than Evolavirus.
13

ICBBE15_Book.indb 13 8/25/2015 7:20:21 AM

 Figure 1. 5 window GP of Ebolavirus amino sequence.

Figure 2. 5 window GP amino sequence of Ebolavirus and Marburgvirus.

Figure 3. 7 window GP amino sequence of Ebolavirus and Marburgvirus.

Figure 5, Figure 6, and Figure 7 are the analysis of VP24 of both viruses under windows 5,
7, and 9. 5-window: Comparison between Ebolavirus amino1 Leucine and Marburgvirus
amino2 Leucine. 7-window: Ebolavirus amino3 Leucine and Marburgvirus amino1 Leucine.
9-window: Ebolavirus amino3 Leucine and Marburgvirus amino3 Leucine. As a result of this
analysis, we concluded that Ebolavirus has higher Leucine in VP24 than Marburgvirus has.
14

ICBBE15_Book.indb 14 8/25/2015 7:20:21 AM

 Figure 4. 9 window GP amino sequence of Ebolavirus and Marburgvirus.

Figure 5. 5 window VP24 amino sequence of Ebolavirus and Marburgvirus.

Figure 6. 7 window VP24 amino sequence of Ebolavirus and Marburgvirus.

Figure 8 and Figure 9 are the analysis of VP35 of both viruses under 7 and 9 windows.
Under 5-windows, there were no amino acids both viruses share to compare with, so we only
chose 7 and 9 windows. 7-window: Comparison between Ebolavirus amino4 Leucine and
Marburgvirus amino5 Leucine. 9-window: Ebolavirus amino6 Leucine and Marburgvirus
amino3 Leucine. In this case, the figures above show that generally Leucine in VP35 of Mar-
burgvirus is higher than that in Ebolavirus.
15

ICBBE15_Book.indb 15 8/25/2015 7:20:22 AM

 Figure 7. 9 window VP24 amino sequence of Ebolavirus and Marburgvirus.

Figure 8. 7 window VP35 amino sequence of Ebolavirus and Marburgvirus.

Figure 9. 9 window VP35 amino sequence of Ebolavirus and Marburgvirus.

4 CONCLUSION

As a result of analysis by Apriori algorithm, we found that the higher the windows are, the
more detailed and diverse the DNAs were. After using Decision Tree, we were able to figure
out similarities between Ebola virus and Marburg virus. Although rules were not extracted
from all windows, it was clear that Ebolavirus and Marburgvirus share similar traits for GP,
16

ICBBE15_Book.indb 16 8/25/2015 7:20:23 AM

 VP24, and VP35. Since, VP24 and VP35 are related with interferon which is used to com-
bat viruses, we are looking forward to find the methods to cure both viruses. For further
research, we want to examine whether having the same hosts leads viruses to possess similar
properties and try to find out the most fitting natural host of Ebolavirus and Marburgvirus
from our studies.

REFERENCES

Henzy, Jamie. 7 Sept. 2014. “Five Questions about Filoviruses.” ‘Small Things Considered’ N.p., Web.
12 Mar. 2015.
Prins, K.C., W.B. Cardenas, and C.F. Basler. “Ebola Virus Protein VP35 Impairs the Function of
Interferon Regulatory Factor-Activating Kinases IKK and TBK-1.” 2009. Journal of Virology 83.7:
3069–077. Web.
Rokach, Lior, and Oded Maimon. 2010. “9. Decision Trees.” Data Mining and Knowledge Discovery
Handbook. N.p.: n.p., n.d. N. pag. Web.
Shi, Zhongzhi. 2011. Advanced Artificial Intelligence. Singapore: World Scientific.
World Health Organization. Sept. 2014. Ebola Virus Disease. WHO, Web. 11 Feb. 2015. Centers for
Disease Control and Prevention. Centers for Disease Control and Prevention, Web. 12 Apr. 2015.
Xu, Wei, Megan R. Edwards, Dominika M. Borek, Alicia R. Feagins, Anuradha Mittal, Joshua B.
Alinger, Kayla N. Berry, Benjamin Yen, Jennifer Hamilton, Tom J. Brett, Rohit V. Pappu, Daisy W.
Leung, Christopher F. Basler, and Gaya K. Amarasinghe. 2014. “Ebola Virus VP24 Targets a Unique
NLS Binding Site on Karyopherin Alpha 5 to Selectively Compete with Nuclear Import of Phospho-
rylated STAT1.” Cell Host & Microbe 16.2: 187–200. Web.

17

ICBBE15_Book.indb 17 8/25/2015 7:20:25 AM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Identifying the missing protein in human proteome

by structure and function prediction

Q.W. Dong
School of Computer Science, Fudan University, Shanghai, China
Department of Computational Medicine and Bioinformatics, University of Michigan,
Ann Arbor, Michigan, USA
Shanghai Key Laboratory of Intelligent Information Processing, Shanghai, China

K. Wang
College of Animal Science and Technology, Jilin Agricultural University, Changchun, China

ABSTRACT: After the completion of human genome project, the proteome research
becomes one of the center problems in post-genomics era. The Human Protein Project aims
to identify at least one protein product from each of the human protein-coding genes by
using experiment methods. However there are still many proteins without experimental evi-
dence which become one of the major challenges in chromosome-centric human proteome
project. Taking into consideration of the complexity of detecting these missing proteins by
using proteomics approach, here we provide the structure and function of these missing pro-
teins by bioinformatics methods. 616 “uncertain” missing proteins are extracted from the
neXtProt database and the structure and function of these missing proteins are predicted
by using state-of-the-art software I-TASSER and COFACTOR respectively. A comprehen-
sive evaluation shows that the results are in good consistent with many manually curated
annotations from well-established databases and other mass spectrum datasets. There are
188 foldable proteins (I-TASSER C-Score larger than −1.5) without using any homologous
template, which may be native gene-coding proteins. The Gene Ontology function predic-
tion results are in good agreement with the manual annotation from neXtProt database,
and also the confidence scores are well correlated with the evaluation metrics with Pearson
correlation coefficient of 0.65. The data are deposited into Human Proteome Structure and
Function database (HPSF) which can provide valuable references about the missing proteins.
The HPSF database is publicly available at http://zhanglab.ccmb.med.umich.edu/HPSF/.

1 INTRODUCTION

Proteins play an important role in biology activities. The successful completion of Human
Genome Project (Venter et al., 2001) provides a valuable blueprint about all the genes encod-
ing entire human proteins. However, due to the complexity of proteins and currently underde-
veloped proteomics technique, many of the proteins have not been detected and annotated.
Towards exploring the universal space of human proteome, the Human Proteome Organi-
zation (HUPO) has recently launched the Human Proteome Project (HPP) (Legrain et al.,
2011) including the Chromosome-centric Human Proteome Project (C-HPP) (Paik et al.,
2012) and Biology/Disease-Driven HPP (B/DHPP). (Aebersold et al., 2013) The primary
goal of the C-HPP is to identify at least one representative protein product and as many post-
translational modifications, splice variant isoforms as possible for each of the human genes.
This ambitious goal is collaborated by 25-membered international consortium covering 24
chromosomes and mitochondria. (Marko-Varga et al., 2013) The HPP executive committee
has established five baseline metrics for C-HPP (Marko-Varga et al., 2013): the Ensembl
database (Flicek et al., 2014) provides the number of protein-coding genes; Peptide Atlas
19

ICBBE15_Book.indb 19 8/25/2015 7:20:25 AM

 (Farrah et al., 2013) and GPMdb (Craig et al., 2004) provide the highly-confident proteins
by mass spectrometry studies; the Human Protein Atlas (Uhlen et al., 2010) provides the
corresponding proteins by antibody-based studies. All the human proteins are deposited in
neXtProt database (Lane et al., 2012) and assigned a confidence code based on the evidence
of the proteins. The proteins from Pe1 level have credible evidence of protein expression
and identification by mass spectrometry, immunohistochemistry, 3D structure, or amino acid
sequencing. The proteins from PE2 level recognize transcript expression evidence, without
evidence of protein expression. The proteins from PE3 level signify the lack of protein or
transcript evidence but the presence of protein evidence for a homologous protein in a related
species. The proteins from PE4 level hypothesize from gene models, and the proteins from
PE5 level come from “dubious” or “uncertain” genes that seemed to have some protein-level
evidence in the past but since has been deemed doubtful.
Recent years, the C-HPP has achieved steady progress. Nearly 78 percent of the total
protein-coding genes have been identified at credible protein evidence by mass spectrometry,
immunohistochemistry, 3D structure, or amino acid sequencing (Lane et al., 2014). However,
there are still more than four thousand proteins which have not been identified by any experi-
mental method and are named as “missing proteins” (Lane et al., 2014) which is constituted
by the proteins from PE2, PE3, PE4 and PE5 level. Most of those missing proteins are hard
to be detected because of the low abundance expression, specific samples, special condition
in biology etc. Identification of the missing proteins is one of the challenges in C-HPP. Many
efforts have been made towards mining the missing proteins by using large-scale mass spec-
trum (Wilhelm et al., 2014, Kim et al., 2014) or bioinformatics methods (Ranganathan et al.,
2013). On the other hand, one of the recent works by Shidhi et al. (Shidhi et al., 2015) aim
to make novel proteins from pseudogenes. By performing a multi-parametric study of the
protein equivalents of the 16 pseudogenes from saccharomyces cerevisiae, they identified two
promising candidates for future protein synthesis in vitro.
This paper will provide comprehensive structure and function analyses of the missing proteins
and try to identify the potential native gene-coding proteins from the missing proteins. Because
of the limited computational resource, currently we only focus on the “dubious” or “uncertain”
(PE5) missing proteins. The structure and function of those missing proteins are predicted by the
state-of-the-art software “I-TASSER” (Roy et al., 2010) and “COFACTOR” (Roy et al., 2012)
respectively. The results are in consistent with the annotation from HGNC database (Gray et al.,
2014) and other mass-spectrum dataset (Kim et al., 2014). All data are publicly available by web-
interface, which can provide valuable reference about the “missing proteins”.

2 METHODS AND PROCEDURES

2.1 Data source of missing proteins

The neXtProt database (Lane et al., 2012) released at Sep. 19, 2014 is used. The “dubious” or
“uncertain” missing proteins with confidence code “PE5” are extracted. Totally, there are 616
proteins in this category with length varying from 21 to 2252. The structures of these proteins
are predicted by using I-TASSER software (Roy et al., 2010). The functions including the EC
number, the GO terms and the binding sites are predicted by using COFACTOR software
(Roy et al., 2012). The subcellular localization is predicted by using Hum-mPLoc (Shen and
Chou, 2009). To validate the presented structure and function prediction method, some of
the high-confidence proteins from neXtProt are also selected with evidence code as “PE1”.
We prefer to the proteins whose 3-D structures are known so that the structure prediction
method can be accessed. The selected proteins have the similar number (625) and length dis-
tribution as the “dubious” missing proteins.

2.2 Pipeline of structure and function prediction for the missing proteins
The procedure for the structure and function prediction is illustrated in Figure 1. For a given
protein sequence, the LOMETS program (Wu and Zhang, 2007) is firstly run to get the possible
20

ICBBE15_Book.indb 20 8/25/2015 7:20:25 AM

 Figure 1. The flowchart of structure and function prediction for missing proteins.

template from PDB library, and then the ThreaDom program (Xue et al., 2013) is used to get
the domain information. If the result is a single domain, the I-TASSER software (Roy et al.,
2010) is used to obtain the final 3-D structure, otherwise the 3-D structure of each domain
is obtained by I-TASSER (Roy et al., 2010), and then these structures are assembled to get a
single structure, which will be refined with FG-MD program (Zhang et al., 2011) to obtain the
final 3-D structure. The COFACTOR program (Roy et al., 2012) is then run to get the function
information including ligand-binding site, Gene-Ontology terms, and Enzyme Classification.
All the programs are extensively tested and achieve good assessment on many community-wide
experiments. For example, I-TASSER was ranked as the No. 1 server for protein structure pre-
diction in recent CASP7 (Moult et al., 2007), CASP8 (Moult et al., 2009), CASP9 (Moult et al.,
2011), and CASP10 (Moult et al., 2014) experiments. The COFACTOR algorithm was ranked
as the best method for function prediction in the CASP9 experiments (Moult et al., 2011).

3 RESULTS AND DISCUSSIONS

3.1 Benchmark test on highly confident protein identification

To give an unbiased evaluation of the proposed method, the structure and function of the
highly confident proteins are predicted in non-homology mode where all the homologous
structures identified with sequence identities greater than or equal to 30% are removed. The
predicted structures are accessed based on the mapping structures from PDB. TM-score (Xu
and Zhang, 2010) is used to quantitatively assess the accuracy of the predicted structure.
Additionally, both I-TASSER (Roy et al., 2010) and COFACTOR (Roy et al., 2012) have
confidence score to indicate the quality of the prediction. The I-TASSER C-score is defined
based on the quality of the threading alignments and the convergence of the I-TASSER’s
structural assembly refinement simulations. These values usually vary from −5 to 2, where a
C-score of higher value signifies a model with a high confidence and vice-versa. An I-TASSER
C-score larger than −1.5 indicates a correct model topology. As shown in Figure 2(A), most
of the highly confident proteins are well-predicted with very high I-TASSER C-scores, and
the average C-score is −0.42. The average TM-score between these highly confident proteins
and the corresponding PDB structures is 0.71, which indicates a high quality prediction. The
Pearson correlation coefficient between I-TASSER C-scores and the TM-score is 0.74, which
means that the I-TASSER C-score is a good indicator of prediction quality.
21

ICBBE15_Book.indb 21 8/25/2015 7:20:25 AM

 Figure 2. The distribution of I-TASSER C-score for PE1 (A) and PE5 (B) proteins respectively.

Figure 3. The distribution of COFACTOR C-score for PE1 (A) and PE5 (B) proteins respectively.

The predicted function is accessed based on the annotation of neXtProt database and
the COFACTOR C-score. The COFACTOR C-score is defined based on the C-score of the
structure prediction and the global and/or local structural similarity between the predicted
models and their structural analogs in the PDB. The COFACTOR C-score has been normal-
ized from 0 to 1, where a high value indicates a high confidence prediction. Among the 625
highly confident proteins, 592 proteins have annotation in neXtProt database and one pro-
tein has no COFACTOR prediction result, so the evaluation is counted on the 591 proteins.
The evaluation metrics used here is the protein-centric metrics as used by Critical Assessment
of Function Annotation (Radivojac et al., 2013), where precision is defined as the number
of correctly predicted functional terms divided by the total number of prediction, the recall
is defined as the number of correctly predicted functional terms divided by the total number
of annotation and F-score is a harmonic mean between precision and recall. As shown in
Figure 3(A), many of the proteins are predicted with high COFACTOR C-score. The Pear-
son correlation coefficient between the COFACTOR C-score and F-score is 0.53, which
means the COFACTOR C-score is also a good indicator of the prediction quality.

3.2 Summary of the predicted structure and function of the missing proteins
To provide the most comprehensive information, the structure and function of the missing pro-
tein are predicted in homology and non-homology mode. In the homology mode, all the possible
templates from PDB library are used no matter they are homologous to the missing proteins
or not. In the non-homology mode, only the non-homologous templates are used, where the
sequence identities between the target missing protein and the templates are below 30%.
22

ICBBE15_Book.indb 22 8/25/2015 7:20:25 AM

 The missing proteins are grouped into three classes based on the result of multi-threading
program LOMETS (Wu and Zhang, 2007). For each threading program, the target-template
alignment is measured by Z-score, which is defined as the difference between the raw alignment
score and the mean in the unit of derivation. A missing protein is classified as “hard” if none
of the threading program identifies a template with Z-score larger than the specific threshold,
a missing protein is classified as “easy” if on average at least one template per threading pro-
gram can be detected with Z-score larger than the specific threshold, otherwise the missing
protein is classified as “medium”. The number of missing proteins in each class is calculated.
As expected, the non-homology mode has more medium and hard target than the homology
mode. Actually, there isn’t a significant difference between the number in homology mode
and that in non-homology mode, which means that our threading programs can efficiently
detect the remote homologous templates. The predicted structure and function in homology
mode may be more accurate than those in non-homology mode, but the results in homology
mode cannot guarantee that the proteins are native gene-coding human proteins. Many of
the missing proteins are inferred from homologous species or derived from the gene model,
so the homologous templates will provide some bias information during prediction. The non-
homology mode doesn’t use any homologous template. If the structures are well predicted in
non-homology mode, the corresponding proteins may be foldable. In this case, the missing
proteins may be native gene-coding proteins.
Because missing proteins have not been validated by any proteomics experiment method,
the native structure and function of these proteins are unknown. Here the confidence scores
of I-TASSER (Roy et al., 2010) and COFACTOR (Roy et al., 2012) are used to access the
quality of the prediction. Figure 2(B) shows the distribution of I-TASSER C-score of PE5
proteins in non-homology mode. The C-score is well-distributed in each bin. The foldable
proteins (with C-score higher than −1.5) are less than the un-foldable proteins (with C-score
lower than −1.5), the reasons may be that the structure of the missing proteins are difficult
to be predicted or some of the missing proteins are not native gene-coding proteins. There
are 188 proteins with I-TASSER C-score larger than −1.5 in non-homology mode (the best
C-score is used for multi-domain proteins). These proteins are foldable without any homolo-
gous information, which indicates that they may be native gene-coding proteins.
The distribution of COFACTOR C-score for PE5 proteins in non-homology mode is
shown in Figure 3(B). The distribution of COFACTOR C-score is more uneven than that of
the I-TASSER C-score. Most values are concentrated between 0.3 and 0.6. There are about
85 proteins with very high COFACTOR. Based on the experience evaluation, the prediction
with COFACTOR C-score higher than 0.2 can hit certain correct GO terms. As shown in
Figure 3(B), most of the missing proteins have good function prediction.

3.3 GO function prediction evaluation

The predicted GO molecular function is mainly enzyme for both missing proteins and highly
confident proteins. The two largest groups are “binding” (GO:0005488) and “receptor activ-
ity” (GO:0003824), which have the frequencies of about 50% and 30% respectively. The miss-
ing proteins are overrepresented with GO items “transporter activity” (GO:0005215) and
“receptor activity” (GO:0004872) in comparison with the highly confident proteins, so it can
be inferred that there are many membrane proteins in the missing proteins.
Among the proteins of PE5 level, there are 219 missing proteins which have function anno-
tations in neXtProt database (Lane et al., 2012). Although these annotations are not golden
standard, they are based on manually curated collection and have high confidence. These
annotations are used to evaluate the performance of our function prediction results. In non-
homology mode, there are 10 proteins whose GO molecular functions have been perfectly
predicted with F-score equal to 1, which means that COFACTOR hits the exact functional
terms. There are 71 proteins with recall equal to 1 at non-homology mode, which means
that COFACTOR can hit the functional terms but wrongly gives more specific function. By
increasing the C-score threshold, COFACTOR can get more accurate prediction as shown
in Table 1. The COFACTOR C-score is well correlated with the prediction results as shown.
23

ICBBE15_Book.indb 23 8/25/2015 7:20:27 AM

 Table 1. The gene ontology molecular function prediction results on 219 missing proteins.

Homology mode Non-homology mode

COFACTOR
C-score No. No.
threshold proteins Precision Recall F-score proteins Precision Recall F-score

0 219 0.42 0.69 0.46 219 0.40 0.64 0.42

0.2 169 0.50 0.72 0.53 174 0.46 0.65 0.47
0.4 139 0.54 0.67 0.56 127 0.53 0.65 0.53
0.6 78 0.76 0.81 0.76 55 0.81 0.87 0.81
0.8 58 0.86 0.86 0.84 42 0.87 0.85 0.84
0.9 39 0.90 0.90 0.88 27 0.91 0.84 0.84

Table 2. Distribution of missing proteins in different gene loci types after HGNC
mapping.

No. foldable proteins

No. missing Homology Non-homology

Gene loci type proteins mode mode

Gene with protein product 66 31 34

Immunoglobulin gene 2 2 2
Pseudogene 252 144 128
RNA, long non-coding 127 2 2
RNA, ribosomal 1 1 0
Withdrawn 6 0 0
Unknown 47 13 12

In homology mode, the Pearson correlation coefficient between the COFACTOR C-score and
F-score is 0.69, where in non-homology mode, the Pearson correlation coefficient is 0.65.

3.4 HGNC mapping analysis

HGNC (Gray et al., 2014) provides a unique name and gene loci type for each of the known
gene within human genome. The majority of HGNC data are manually curated (Gray et al.,
2014). The gene loci type is valuable information to verify the missing proteins. Within the
total 616 missing proteins here, there are 507 proteins which can be mapped to one or more
HGNC ID. Table 2 shows the possible gene loci types, the number of missing proteins in
each loci and the number of foldable proteins (with I-TASSER C-score larger than −1.5).
As expected, the number of foldable proteins in homology mode is more than that in non-
homology mode except the gene with protein product class. Actually, the reason to this
exception is that the target proteins are medium targets. Half of the proteins in gene with
protein product class have been identified as foldable proteins. The others may be hard to
be predicted. There are many foldable proteins from the pseudogene class. Pseudogenes are
dysfunctional relatives of genes that have lost their protein-coding ability or are otherwise no
longer expressed in the cell. Usually, these pseudogenes have homologous protein products
during evolution, so I-TASSER can fold many of the proteins in this class.

3.5 Comparison with other mass spectrometry dataset

Mass spectrometry is an effective method to identify peptides. Lots of mass spectrometry
data has been deposited in public database, such as PeptideAtlas (Desiere et al., 2006) and
GPMDB (Craig et al., 2004). Several groups try to use mass spectrometry to develop the
draft of human proteome (Wilhelm et al., 2014, Kim et al., 2014). One of the interesting
results reported by Kim et al (Kim et al., 2014) is that they have identified about two-third
24

ICBBE15_Book.indb 24 8/25/2015 7:20:27 AM

 (2535/3844) of the “missing proteins” (Lane et al., 2014). Actually the “missing proteins” in
that paper refers to the neXtProt proteins whose evidence codes are “PE2”, “PE3” or “PE4”.
This paper focuses on the “PE5” missing proteins. By RefSeq (Pruitt et al., 2014) mapping, 41
“PE5” proteins are also found in Kim’s dataset. In non-homology mode, 23 out of the 41 pro-
teins are foldable (I-TASSER C-score>−1.5), which is in good consistent with Kim’s results.

3.6 Web interface

The predicted structure and function of the 616 missing proteins are publicly available at
http://zhanglab.ccmb.med.umich.edu/HPSF/. User can browse or search the specific pro-
teins by clicking the “Browse & Search” link. The proteins can be searched by inputting the
neXtProt ID, protein name and corresponding HGNC symbol or name. Both partial and full
values are accepted. The input box can automatically provide at most 20 suggestions if any
of the record can match the inputted text. The browse or search results page displays general
information of the proteins, including the neXtProt ID, the protein name, the gene name, the
gene loci type, the ENSEMBL gene, the prediction confidence score (I-TASSER C-score)
et al. User can click the link to access the detail information provided by other databases,
including neXtProt (clicking neXtProt ID), HGNC (clicking gene symbol) and ENSEMBL
(clicking the ensemble gene). To get the detail structure and function information, please click
the “HPSF detail” link under the neXtProt ID which will open a new window. The new page
will show comprehensive predicted structure and function information about the missing
proteins. The structure information includes the predicted secondary structure, the predicted
solvent accessibility, the predicted B-factor, the top 5 models and their residue-specific quali-
ties and the domain information. The function information includes the Enzyme commission
number (Barrett, 1997) and Gene Ontology terms (Ashburner et al., 2000). The page also
give the predicted subcellular localization by Hum-mPLoc (Shen and Chou, 2009). For each
missing proteins, the similar proteins from known proteins (PE1 level) are also extracted and
displayed. User can also change the results between the homology mode and non-homology
mode by simply clicking the corresponding link. The results of each domain can also be
accessed by clicking the link of each domain. Finally, the whole database can be downloaded
at the download page.

4 CONCLUSION

In this study, 616 missing proteins, which have the lowest confidence (PE5) in neXtProt database,
have been investigated by using bioinformatics methods. The structure and function of these
proteins have been predicted by using cutting-edge software: I-TASSER and COFACTOR
respectively. The prediction is extensively evaluated by extracting well-established evidence
about missing proteins, such as neXtProt annotation, HGNC gene loci annotation and mass-
spectrometry dataset. The results show that there is good consistency between the prediction
and the evidence of the proteins in PE5. There are 188 foldable proteins with high confidence
score of I-TASSER structure simulation without using any homologous templates, indicating
that these proteins may be native gene-coding proteins. Both structure and function evalua-
tion shows that the missing proteins are over-represented in membrane proteins in comparison
with the highly confident proteins. Since the membrane proteins are hard to be separated
and purified, detection of the membrane proteins are more difficult than other proteins. The
results indicate that there may be more membrane proteins in the missing proteins.

REFERENCES

Aebersold, R., Bader, G.D., Edwards, A.M., et al. 2013. The biology/disease-driven human proteome
project (B/D-HPP): enabling protein research for the life sciences community. J Proteome Res, 12, 23–7.
Ashburner, M., Ball, C.A., Blake, J.A., et al. 2000. Gene ontology: tool for the unification of biology.
The Gene Ontology Consortium. Nat Genet, 25, 25–9.

25

ICBBE15_Book.indb 25 8/25/2015 7:20:27 AM

 Barrett, A.J. 1997. Nomenclature Committee of the International Union of Biochemistry and Molecular
Biology (NC-IUBMB). Enzyme Nomenclature. Recommendations 1992. Supplement 4: corrections
and additions (1997). Eur J Biochem, 250, 1–6.
Craig, R., Cortens, J.P., Beavis, R.C. 2004. Open source system for analyzing, validating, and storing
protein identification data. J Proteome Res, 3, 1234–42.
Desiere, F., Deutsch, E.W., King, N.L., et al. 2006. The PeptideAtlas project. Nucleic Acids Res, 34,
D655–8.
Farrah, T., Deutsch, E.W., Hoopmann, M.R., et al. 2013. The state of the human proteome in 2012 as
viewed through PeptideAtlas. J Proteome Res, 12, 162–71.
Flicek, P., Amode, M.R., Barrell, D., et al. 2014. Ensembl 2014. Nucleic Acids Res, 42, D749–55.
Gray, K.A., Yates, B., Seal, R.L., et al. 2014. Genenames.org: the HGNC resources in 2015. Nucleic
Acids Res.
Kim, M.S., Pinto, S.M., Getnet, D., et al. 2014. A draft map of the human proteome. Nature, 509, 575–81.
Lane, L., Argoud-Puy, G., Britan, A., et al. 2012. neXtProt: a knowledge platform for human proteins.
Nucleic Acids Res, 40, D76–83.
Lane, L., Bairoch, A., Beavis, R.C., et al. 2014. Metrics for the Human Proteome Project 2013–2014 and
strategies for finding missing proteins. J Proteome Res, 13, 15–20.
Legrain, P., Aebersold, R., Archakov, A., et al. 2011. The human proteome project: current state and
future direction. Mol Cell Proteomics, 10, M111 009993.
Marko-Varga, G., Omenn, G.S., Paik, Y.K., et al. 2013. A first step toward completion of a genome-
wide characterization of the human proteome. J Proteome Res, 12, 1–5.
Moult, J., Fidelis, K., Kryshtafovych, A., et al. 2007. Critical assessment of methods of protein structure
prediction-Round VII. Proteins, 69 Suppl 8, 3–9.
Moult, J., Fidelis, K., Kryshtafovych, A., et al. 2009. Critical assessment of methods of protein structure
prediction—Round VIII. Proteins, 77 Suppl 9, 1–4.
Moult, J., Fidelis, K., Kryshtafovych, A., et al. 2014. Critical assessment of methods of protein structure
prediction (CASP)—round x. Proteins, 82 Suppl 2, 1–6.
Moult, J., Fidelis, K., Kryshtafovych, A., et al. 2011. Critical assessment of methods of protein structure
prediction (CASP)—round IX. Proteins, 79 Suppl 10, 1–5.
Paik, Y.K., Jeong, S.K., Omenn, G.S., et al. 2012. The Chromosome-Centric Human Proteome Project
for cataloging proteins encoded in the genome. Nat Biotechnol, 30, 221–3.
Pruitt, K.D., Brown, G.R., Hiatt, S.M., et al. 2014. RefSeq: an update on mammalian reference
sequences. Nucleic Acids Res, 42, D756–63.
Radivojac, P., Clark, W.T., Oron, T.R., et al. 2013. A large-scale evaluation of computational protein
function prediction. Nat Methods, 10, 221–7.
Ranganathan, S., Khan, J.M., Garg, G., et al. 2013. Functional annotation of the human chromosome
7 “missing” proteins: a bioinformatics approach. J Proteome Res, 12, 2504–10.
Roy, A., Kucukural, A., Zhang, Y. 2010. I-TASSER: a unified platform for automated protein structure
and function prediction. Nat Protoc, 5, 725–38.
Roy, A., Yang, J., Zhang, Y. 2012. COFACTOR: an accurate comparative algorithm for structure-based
protein function annotation. Nucleic Acids Res, 40, W471–7.
Shen, H.B., Chou, K.C. 2009. A top-down approach to enhance the power of predicting human protein
subcellular localization: Hum-mPLoc 2.0. Anal Biochem, 394, 269–74.
Shidhi, P.R., Suravajhala, P., Nayeema, A., et al. 2015. Making novel proteins from pseudogenes.
Bioinformatics, 31, 33–9.
Uhlen, M., Oksvold, P., Fagerberg, L., et al. 2010. Towards a knowledge-based Human Protein Atlas.
Nat Biotechnol, 28, 1248–50.
Venter, J.C., Adams, M.D., Myers, E.W., et al. 2001. The sequence of the human genome. Science, 291,
1304–51.
Wilhelm, M., Schlegl, J., Hahne, H., et al. 2014. Mass-spectrometry-based draft of the human proteome.
Nature, 509, 582–7.
Wu, S., Zhang, Y. 2007. LOMETS: a local meta-threading-server for protein structure prediction.
Nucleic Acids Res, 35, 3375–82.
Xu, J., Zhang, Y. 2010. How significant is a protein structure similarity with TM-score = 0.5? Bioinfor-
matics, 26, 889–895.
Xue, Z., Xu, D., Wang, Y., et al. 2013. ThreaDom: extracting protein domain boundary information
from multiple threading alignments. Bioinformatics, 29, i247–56.
Zhang, J., Liang, Y., Zhang, Y. 2011. Atomic-level protein structure refinement using fragment-guided
molecular dynamics conformation sampling. Structure, 19, 1784–95.

26

ICBBE15_Book.indb 26 8/25/2015 7:20:27 AM

 Bioinformatics and Biomedical Engineering – Chou & Zhou (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02784-8

Prediction of SARS coronavirus main protease

by artificial Neural Network

S.P. Jang, S.H. Lee, S.M. Choi, H.S. Choi & T.S. Yoon
Natural Science, Hankuk Academy of Foreign Studies, HAFS, Yongin-si, Republic of Korea

ABSTRACT: SARS (Severe Acute Respiratory Syndrome) corona virus has hugely affected
humans for more than ten years. Virus’ RNA replicase gene is surrounding the polyprotein
la and lab, and the sequence of the polyproteins contains functional proteins, which is an
important factor of replication. The experiment was performed based on the “distorted key
theory” in order to prevent SARS corona virus from performing replication by inactivating
the main protease (also called CoV Mpro) of the virus. After the experiment, Neural Net-
work (NN) was utilized in order to reanalyze the results of polypeptides in the virus. This
approach by NN distinguished the fixed patterns in the sequence of cleavage site successfully,
and improved the comprehension of the protease structure. The method of preventing the
virus replication using competitive inhibitor could be found by analyzing these patterns.

1 INTRODUCTION

Severe Acute Respiratory Syndrome (SARS) is a respiratory disease, which has largely influ-
enced both humans and animals. After the outbreak of SARS in Asia, the WHO declared
the coronavirus, which is classified as a single-strand RNA virus of zoonotic origin, as the
main cause of SARS. Between November 2002 and July 2003, SARS had a serious effect on
8,273 individuals and caused 775 deaths (9.6% mortality rate) in multiple countries, with most
cases in Hong Kong (Chou K.C. 1996). The initial symptom of the sickness was high fever
above 38°C (100.4 °F) with an unspecific flu-like symptom, involving breathing difficulties.
SARS coronavirus main protease (CoV Mpro) is an enzyme that catalyzes RNA replicase of
the virus, which is an essential process for the virus to survive, through replicating essential
polyprotein (Marra Marco A. et al. 2003). Perceiving the virus to be disastrous to humans,
we decided to perform a study on the cleavage site in CoV Mpro, which is known as the secret
of developing drugs against SARS because of its status as an intimidator of SARS existence.
Based on the “distorted key theory”, we performed an experiment using the NN algorithm to
analyze the sequence of the cleavage site and increase its analytic accuracy in order to develop
an effective amino acid, which is a key factor to prevent viral replication.

1.1 Neural Network

Deriving the usefulness of the human central nervous system, Neural Network (NN) is used
to work out with numerous cases of problems mathematically. Similar to the biological nerv-
ous system, there are ‘neurons’—simple artificial nodes—which are attached to each other
in order to construct a network (Hansen Lars Kai & Peter Salamon 1990). Every artificial
neuron works as an individual processor, which enables the total network system not to be
affected by errors in some neurons or in other words ‘fault tolerance’. Furthermore, NN
is able to develop itself through the given situations. This makes it to be used widely in the
problem solving of Artificial Intelligence (AI), voice recognition, character recognition and
in other diverse fields. As a way to acquire the efficient inhibitor sequence of the protease and
assume the structure of protein, we used the NN algorithm. For the input, we inserted amino

27

ICBBE15_Book.indb 27 8/25/2015 7:20:27 AM

 acid sequences in order to analyze the ones that appear the most. These have high possibility
to obtain an effective inhibitor sequence. By using NN as a way of analyzing, we were able to
predict more accurately than C5, decision tree and other methods.

1.2 Distorted key theory

An useful method of developing inhibitors against SARS is distorted key theory. Effective
information for finding inhibitors that are against the key enzyme can be known by protein

Figure 1. Each circle represents individual nodes that are connected together, arrows represent the
output for one side and the input for another side.

Figure 2. Chou’s distorted key theory: (a) the peptide both cleaved by CoV Mpro and effectively
bound to the active site of the protease, while the peptide in plate (b) becomes non-cleavable through
modification but still bound to the active site. The modified peptide, also called a “distorted key”, can
become an effective inhibitor against SARS.

28

ICBBE15_Book.indb 28 8/25/2015 7:20:27 AM

 cleavage sites. As Koshland’s induced fit theory and Fisher’s lock-and-key model propose, for
the CoV Mpro to cleave a peptide with high possibility, the binding between the active site
of the enzyme and the substrate is vital (Zheng Rong Yang 2005). However, once a peptide’s
scissile bond is modified to a strong hybrid chemical bond, and even though it is still able to
bind to the active site, its probability of cleaving would plunge. From this, the peptide can be
regarded as the “distorted key”, in a way that it can be inserted in the lock but can neither
unlock nor come out. As a result of strong binding of the modified peptide, it can operate as
an efficient inhibitor against the enzyme (Qi-Shi Du et al. 2007).

2 METHOD AND EXPERIMENT

Neural Network (NN), as the name suggests, is a computer program imitating the informa-
tion procedure of the biological nervous system. The nervous system consists of basic units
called neurons. Neurons are linked to each other, developing a huge network as a whole. The
processing of an individual neuron is known as follows: a new electrical signal is received
through the dendrites, transferred down the axon, and finally sent to other neighbored
neurons. The process of receiving and passing the signal can be carried out only when the
signal exceeds the given threshold. Also, a single neuron can be linked with multiple neurons
at both ends.
Obviously, NN is a set of nodes in a same topology linked to each other. Each node has
input and output links to the others. The fact that these links allow every node to interact
with the others and the output value depends on whether the signal surpasses the threshold
corresponds to the processing in biological neurons. Reflecting this correspondence, the main
function in NN that actually processes the given information is called “neurons.”

2.1 Artificial Neural Network

The NN (Neural Network) information procedure can be explained by how perceptrons work
(perceptron is a pattern classification device which has learning ability). NN is constructed
by several numbers of perceptrons and it works by their cooperative operation. Perception
is a one direction, as shown in Figure 1. Several inputs are given and in the next stage each
input is multiplied with specific weight factors that range from 0 to 1. After this process, each
product value (x*w) is added and the whole sum (∑x*w) will be sent to the neuron. If the sum
is higher than the threshold, the signal results in the output, otherwise no signal is sent out.
The experiment output is determined by the NN learning (NN learning is done before the
experiment). The learning operation is possible when the input and the output values are
designed previously. Let us introduce the progress of the NN learning. First, random weight
factors are given to the input value. Second, the result of the first step is compared with the
output value that we are trying to analyze. Third, the weight factor multiplied with each of
the input value is modified to match the real put with the desired one. After these operations,
neurons are able to be classified newly by the NN learning progress.

Figure 3. Perceptron is the principle of NN (Neural Network), calculating the product of the input
and weight values and their sum, from which the result can be obtained.

29

ICBBE15_Book.indb 29 8/25/2015 7:20:31 AM

 Table 1. Results of 8-mers.

Site Most frequent Second-most Third-most

R4 A: 36.4% V: 27.3% P: 18.2%

R3 T: 36.4% K: 18.2% R: 18.2%
R2 L: 72.7% F: 9.1% M: 9.1%
R1 Q: 100% – –

In our experiment, we marked 20 amino acids with numbers ranging from 1 to 20. Next,
these numbers were given as inputs to the NN. The input product value and the weight fac-
tor were put into the neuron, and the output value is earned by the previously designed NN.
Analyzing the frequency of the certain amino acids in a specific spot, it was possible to know
the sequence patterns.
The polyprotein, which is related to CoV Mpro, can be parted into two groups: 8-mers and
12-mers. The rules shown between the two groups were considerably different: NC_004718
(TOR2), NC_002645 (HCoV 229E), NC_001846 (MHV), NC_003045 (BCoV), NC_--1451
(IBV), NC_002306 (TGEV), NC_003436 (PEDV), AF208067 (MHVM), AF201929 (MHV2),
AF208066 (MHVP), AY278741 (Urbani), AY278488 (BJ01), AY278554 (CUHK-W1),
AY282752 (CUHK-su10), and AY291451 (TW1). The experiment was performed with the data
of 154 8-mers and 45 12-mers within 14 cleavage sites. Finally, we gained the data from ‘Mining
SARS-CoV protease cleavage data using non-orthogonal decision trees, a novel method for
decisive template selection’, researched by Zheng Rong Yang (Zheng Rong Yang 2005).

3 RESULTS

In NN results, we could find a strong rule in the 8-mers experiment, but we could not find a
special rule because the 12-mers experiment did not show high accuracy and strong rule. The
results for 8-mers are given in Table 1.

3.1 8-mers
The amino acids of performing neural network algorithm with the 8-mers are as follows:
amino1 in {A,I,K,N,P,S,T,V}
amino2 in {A,D,F,G,I,K,L,M,N,Q,R,S,T,V,B}
amino3 in {F,I,L,M,V}
amino4 in {Q}
amino5 in {A,C,G,N,S}
amino6 in {A,C,E,F,G,I,K,L,N,R,S,T,V}
amino7 in {A,D,E,F,G,I,K,L,M,N,P,Q,R,S,T,W,Y}
amino8 in {A,D,E,G,H,I,K,L,M,N,P,Q,R,S,T,V,W}

3.2 12-mers
Unexpectedly, no key rule was found in the performance involving the 12-mers. The presence
of different amino acids among all the sites did not have differences that are huge enough to
consider as the key factor.

4 CONCLUSIONS

This paper applied the computational program, NN (Neural Network). NN is composed

of interaction between perceptrons. NN produces the sequence patterns of cleavage sites by
30

ICBBE15_Book.indb 30 8/25/2015 7:20:32 AM

 reanalyzing the polypeptides of SARS coronavirus. This outcome leads to representing the
effective inhibitor, which induces the inactivation of CoV Mpro based on the “distorted key”
theory.
The SARS virus is a very critical and epidemic disease. In order to block the virus from
spreading out and killing people, we performed the experiment based on the distorted key
theory and the neural network algorithm.
The rule extracted from the 8-mers neural network showed that amino acid Glutamine is
the main factor of CoV Mpro. However, we could not find new rules in the 12-mers result.
In our experiment, we found that A, V, T, K, L, Q are key factors that make CoV Mpro to
operate. In related work, another experiment that analyzed CoV Mpro performing Support
Vector Machine showed the amino sequence A-V-L-Q-S-G-F-R (Shinyoung Lee et al. 2014).
In conjunction with the above result, our experiment showed a very accurate result.

REFERENCES

Chou K.C. (1996). “Review: Prediction of human immunodeficiency virus protease cleavage sites in
proteins”. Analytical Biochemistry 233: 1–14. doi:10.1006/abio.1996.0001. PMID 8789141. (references).
Hansen, Lars Kai, and Peter Salamon. “Neural network ensembles.” IEEE transactions on pattern
analysis and machine intelligence 12.10 (1990): 993–1001.
Marra, Marco A., et al. “The genome sequence of the SARS-associated corona-virus.” Science 300.5624
(2003): 1399–1404.
Qi-Shi Du, Hao Sun and Kuo-Chen Chou, “Inhibitor Design for SARS Coronavirus Main Protease
Based on Distorted Key Theory,” Medicinal Chemistry, 2007, 3, 1–6.
Shinyoung Lee, Jisue Kang, Jiwoo Oh, Yoonjoo Kim, Jungwon Baek and Taeseon Yoon, “Prediction of
SARS Coronavirus Main Protease by Support Vector Machine”, IACSIT Press Vol. 59, 2014.
Zheng Rong Yang, “Mining SARS-CoV protease cleavage data using non-orthogonal decision trees;
a novel method for decisive template selection,” Vol. 21 no. 11 2005, pages 2644–2650. doi:10.1093/
bioinformatics/bti404.
Zheng Rong Yang, “Mining SARS-CoV protease cleavage data using non-orthogonal decision trees;
a novel method for decisive template selection,” Vol. 21 no. 11 2005, pages 2644–2650. doi:10.1093/
bioinformatics/bti404.

31

ICBBE15_Book.indb 31 8/25/2015 7:20:32 AM

Another Random Scribd Document
with Unrelated Content
 Die Strasse führt im Thalzug weiter nach Schmiedeberg. Hohofen.
Eisenhammer. Hinter dem Ort zeigen weisse T, links der Strasse, nach
der Tellkoppe hinauf, dieselbe ist jedoch jetzt so verwachsen, dass
sich eine Ersteigung erst dann wieder lohnt, wenn das projektirte
Ausschaugerüst fertig gestellt sein wird. Die 754 m hohe Kuppe auf
langem, waldigen Bergrücken gewährte umfassende Blicke ins
Centralerzgebirge hinauf und auf die Höhenzüge bei Freiberg, wie auf
das Meissner Tiefland. Wir gehen besser direct nach der
Friedrichshöhe bei Bärenburg. 739 m. An einer grösseren Tanne ist
die Aussicht am schönsten. Wir sehen Dresden mit den neuen
Kasernen, die Prinzenschlösser, Theile der Sächs. Schweiz, die Berge
bei Altenberg, die Lausitzer Berge und die nördliche Abdachung des
östl. Erzgebirges selbst. König Friedrich August nächtigte auf seinen
botanischen Wanderzügen öfter hier und gab Raketensignale hinab
nach Pillnitz, wo seine Gemahlin dieselben erwidern liess. In der
Försterei intr. Königszimmer mit Fremdenbuch für fürstl. Personen. Die
Strasse führt von hier über Hirschsprung nach Altenberg. Siehe Seite
32.

6. Dresden-Dippoldiswalde-Schmiedeberg
(26 km). Rehefeld-Zaunhaus (über Dorf Sayda
14 km, über Bärenfels und Schellerhau 13 km, doch
mehr Berg). Neustadt (6 km). Niclasberg (3 km).
Klostergrab (5 km). Ossegg (4 km). Teplitz
(10 km).
Bis Schmiedeberg siehe Tour 5. Hier verlassen wir die rothe Weisseritz
und gehen im Pöbelbachthal aufwärts die Strasse, die bei Ober-Pöbel
einen Bergrücken überschreitet und das Thal der Wilden Weisseritz
gewinnt. In Sayda Gasthof. Die Strasse folgt in der Hauptsache der
Weisseritz bis Rehefeld-Zaunhaus.

Anmerkung. Wer von Schmiedeberg aus den Weg über Bärenfels

und Schellerhau wählt, kann bei letzterem einsamen Gebirgsdorf den
842 m hohen Pöbelknochen ersteigen. Umfassender Blick in die
einsame Waldregion des Kammes und auf dem Nordabhang des
Gebirges. Die Strasse führt dicht daran vorüber.
 Rehefeld-Zaunhaus. Jetzt eine Gemeinde. Zerstreutes
Gebirgsdorf mit einem königl. Jagdhaus, ein Lieblingsort König
Alberts und seiner Gemahlin. Liegt inmitten herrlicher Waldberge an
der Wilden Weisseritz, dicht an der böhmischen Grenze. Primitiver
Landgasthof. Wenn der Hof anwesend, ist oft schwer Unterkommen
zu finden, so dass sich selbst Cavaliere mit Strohlagern in
Bauerhütten begnügen müssen. Zeichnung und Bauproject zu dem
hölzernen, doch höchst malerischen Jagdschlösschen bescheerte die
Königin Carola ihrem erlauchten Gemahl als ein
Weihnachtsgeschenk. Schöne Kapelle in norwegischer
Holzarchitektur. Glasgemälde. Gutes Altarbild. Die Einrichtung des
Schlösschens ist von einer fast rührenden Einfachheit. In der
Umgebung unter sehr alten Bäumen schöne Waldprommenaden. Der
Hemmschuhberg mit sehr hohem Holzgerüst zwischen drei Fichten
gestattet umfassende Blicke in die Waldregion des Kammes (815 m).
Wegweiser markiren die herrlichen Waldwege nach dem Thurm
(4 km).
Ein directer Weg (6 km) führt über das Forsthaus Kalkofen nach
Niclasberg hinab. Auf der Höhe (an einem Kreuz) überraschende
Aussicht durch den Niclasberger Thalzug hindurch auf das Mittelgebirge
und den Teplitzer Thalkessel. Wer den Stürmer nicht ersteigen will,
gehe diesen directen Weg. Andernfalls berühren wir Neustadt.
Städtchen auf der Kammhöhe. Schütz' Gasthaus. Biliner Bierhalle. Von
hier liegt der Stürmer nur 1½ km und ist leicht zu ersteigen, doch wird
man ohne Führer die aussichtsreiche Kuppe nur schwer finden. Der 825
m hohe, weitvorgeschobene Berg gehört zu den grossartigsten
Aussichtspunkten des Erzgeb. Die Ausschau auf den Teplitzer
Thalkessel wie auf das bizarre Mittelgebirge ist ausserordentlich
umfassend, auch die Ebene gegen Saaz hin liegt frei vor dem
Beschauer.
Von Neustadt führt die aussichtsreiche Strasse, vom Stürmer führen
directe Waldpfade steil hinab nach Niclasberg oder auch direct nach
Klostergrab. Fall bis dahin nahezu 500 m. Das kleine Bergstädtchen
liegt höchst anmuthig im engen Thalzuge. Rathskeller am Markt (Gut).
Das Städtchen eignet sich sehr wohl zur Sommerfrische.

Anmerkung. Am Weg von Niclasberg nach Eichwald (6 km) liegt

rechts der Strasse der Wolfsstein, ein prächtiger Aussichtspunkt,
nicht hoch, aber äusserst günstig für Ausblicke auf Teplitz und das
Mittelgebirge.
Die Strasse von Niclasberg hinab nach Klostergrab führt durch eine enge
romantische Waldschlucht.

Klostergrab. Zum Rathskeller. 1400 Einw. 353 m. Die Stadt ist

geschichtlich merkwürdig wegen einer protestantischen Kirche, die
der Prager Erzbischof von Prag Lohelius 1616 niederreissen liess.
Dadurch kam der Zündstoff, der sich im religiösen Leben
Deutschlands seit der Reformation aufgehäuft, zum Brennen. Die
protest. Landstände Böhmens, darunter ein Lobkowitz, ein Schick
und ein Kinsky, forderten in Prag Rechenschaft, was bekanntlich
Veranlassung ward, zwei Kaiserl. Räthe aus dem Fenster zu stürzen,
worauf der 30-jähr. Krieg entbrannte. Die Ruinen dieser Kirche, die
am Rathhaus stand, sind völlig beseitigt. 60 m über der Stadt liegt
der Königshügel mit Rest. Herrliche Aussicht auf den Teplitzer
Thalkessel und das Mittelgebirge.
Nach Ossegg führt Eisenbahn; links sitzen. Ununterbrochen schöne
Aussicht. Zu Fuss geht man über Haan. Unterwegs ein auffälliger, sehr
grosser Eichbaum.

Ossegg. Marktflecken mit 2000 Einw. Kaiser v. Oesterreich.

Sonne. Villencolonie. Der malerische Ort liegt mit seinem
grossartigen Kloster am Fusse des imposanten Strobnitzberges. Das
Kloster gehört dem Cisterzienserorden; 1196 unter dem Schutze
der Riesenburg gegründet, ward es in den Hussitenkriegen
mehrmals völlig ausgeraubt. Rudolph II. hob es im 16. Jahrh. ganz
auf, doch gelangte es nach Unterdrückung der Reformation im 17.
Jahrh. wieder zu Reichthum und Blüthe. Es leben im Kloster etwa 20
Mönche. (Schwarz-weisse Ordenstracht.) Der Abt ist auch Prälat der
Sächs. Nonnenklöster Marienthal und Marienstern.
Der Klostergarten ist im Versailler Styl mit Schnitthecken,
Orangerien, Treibhäusern und Bassins angelegt. Sehr schöne
Aussichtspunkte von den Galerien. Zutritt frei.
Die Stiftskirche, ein Renaissancebau mit sehr reicher
Ausstattung. Grosse herrliche Orgel. Trefflich stylisirter Altar, schön
geschnitzte Chorstühle, reich ornamentirte Leuchter, Gemälde,
Stukkaturarbeiten. Die anstossende Todtenkapelle wird magisch von
oben beleuchtet. Der Kreuzgang, aus den besten Zeiten der Gothik
liegt innerhalb der Clausur und ist daher für Damen nicht zugänglich,
dasselbe gilt auch vom Conventsgebäude mit dem Capitelsaal. Im
Refectorium (30 kr. Trinkgeld) hat man herrliche Aussicht auf das
Dux-Teplitzer Thal und das Mittelgebirge. Kunstvoll geschnitzte
Eckschränke, alte Kamine. In den nahen Gasträumen Bildergalerie,
darunter zwei treffliche Portraits von Ossegger Aebten. Viele Copien
von Meisterwerken.
Die Bibliothek, über deren Eingang die Worte stehen: »Non pro
spectaculo, sed pro usu«, enthält seltene Werke und Manuscripte.
Die Salesiushöhe 2 km vom Kloster, ist ein zugänglich
gemachtes Gewirr von Felsblöcken mit herrlicher Aussicht auf den
Thalkessel und das Mittelgebirge, doch finden sich gleich schöne
Punkte auf den näheren Bergen im Hintergrunde von Ossegg.
Die Riesenburg erhebt sich 3½ km entfernt im Thalzug hinter
Ossegg. Siehe Seite 45.
Die Stadt Dux liegt 4½ km von Ossegg. Die Strasse führt durch die
bekannten inundirten Ossegger Schächte (Dux s. S. 25). Direct nach
Teplitz geht man von Ossegg über das intr. Dorf Janegg (s. S. 25).
Auch führen zwei Bahnlinien dahin. Vom Bahnhof der Dux-Bodenbacher
Bahn (Waldthorbahnhof) hat man freilich noch 3 km nach Teplitz zu
gehen. Teplitz siehe Seite 21.

7. Dresden-Tharandt (15 km). Edle Krone

(4 km). Höckendorf (2½ km). Beerwalder Mühle
(6½ km). Frauenstein (9½ km). Georgensdorf
(13 km). Fley (7 km). Langewiese (5½ km).
Ossegg (über Riesenburg 5 km). Teplitz (10 km).
Man benutzt am Besten bis Tharandt die Eisenbahn, die sich durch den
industriereichen Plauenschen Grund und später ab Heinsberg durch ein
 schönes Waldthal hinzieht.

Tharandt. Deutsch. Haus. Erblehngericht. Albertsalon und Bad,

beide innerhalb schöner Gärten. Burgkeller unter der Ruine. 2600
Einw. Tharandt ist ein liebliches kleinstädtisches Idyll an einem Punkt
gelegen, an dem zwei Thäler zusammenstossen. Auf einer Bergnase,
die sich vom Forstgartenberg abhebt, liegt die Ruine eines alten
Jagdschlosses und das anmuthige Kirchlein. Tharandt war ein
Lieblingsaufenthalt Albrecht des Beherzten, dem Erbauer der
Albrechtsburg. 1668 ward das Jagdschloss vom Blitz eingeäschert
und liegt seitdem in Trümmern. Die Forstakademie, unter den
Ruinen gelegen, ward 1816 durch Cotta, den berühmten Forstwirth,
begründet (im Durchschnitt 130 Studirende, ⅔ Ausländer). Die
Akademie besitzt fachwissenschaftliche Sammlungen, die auf
Ersuchen zugänglich sind. Von besonderem Interesse sind die
Querdurchschnitte zweier Riesenfichten, aus erzgeb. Forsten
herrührend (lagern im Hausflur). Ueber künstliche Fischzucht wird
alljährlich im November von Prof. Krutzsch ein practischer Cursus
unentgeltlich abgehalten. Der Forstgarten ist eine sehr
ausgedehnte Pflanzung für Lehrzwecke mit allen Laub- und
Nadelhölzern, die in unseren Gegenden gedeihen. Der Königsplatz
mit herrlichen Thalblicken und das Cottagrab, beide im Forstgarten,
verdienen besucht zu werden. 80 Eichen symbolisiren die 80
Lebensjahre Cotta's. Bemerkt sei, dass der Forstgarten nicht immer
offen steht. Einem Schmuckkästchen gleicht das gräfl.
Sumynsky'sche Schlösschen mit einem geschmackvoll angelegten
und peinlich gepflegten Garten. Im Bad jährlich gegen 400 Kurgäste.
Oefter Concerte. Starker Fremdenbesuch. Tharandt mit seinen
Buchenbeständen an imposanten Berghängen ist im Frühjahr und
Herbst unvergleichlich schön.
Die Thalmühle, 3 km vom Städtchen entfernt, ist eine
vielbesuchte idyllisch-primitive Bauernschenke. Schöne Promenaden
führen auch an beiden Thalhängen abwärts nach Hainsberg. Die
Heiligen Hallen (ein schöner Buchenforst) liegen aufwärts nach
Edle Krone zu rechts.
 Wir wandern nun im schönen Thalzug neben Bahn und Weisseritzfluss
oder durch die Heiligen Hallen aufwärts nach Edle Krone.
Bahnstation. Gasthof. Hier wenden wir uns links in das romantische
Höckenbachthal. Inschriften, darunter eine solche an der sogenannten
Wunderbuche folgenden Wortlautes:

Aus der Wunderbuche schallt Vögleins Klage laut

erhoben:
Wer hat dich, du schöner Wald, abgeholzt so hoch dort
oben?

Höckendorf. 1100 Einw. Weckbrods Wein- und Bierschank. Die

Kirche enthält einen geschnitzten Altarschrein, den Kenner
künstlerisch sehr hoch stellen. Zwischen dem alten und dem neuen
Friedhof steht eine von den drei uralten Martersäulen, welche ein
Ritter Conrad von Theeler setzen liess, nachdem er im Zorn seinen
Pfarrer erstochen. Höckendorf trieb früher starken Bergbau, dieser
wurde aber am 25. Aug. 1557 durch einen Wolkenbruch plötzlich
vernichtet.
In Höckendorf geht die »Butterstrasse« direct nach der Beerwalder Mühle
ab, doch ist die Strasse, die ihren Namen von den erzgeb.
Buttertransporten nach Dresden erhielt, ohne Specialkarte nicht zu
finden, auch ist der Umweg über Ruppendorf nur unbedeutend. In
Ruppendorf steht neben dem Gasthof flach auf einer Wiese eine
Thurmruine, die der Volkssage nach ein Raubschloss war, davon soll
das Dorf (Raubdorf) seinen Namen haben. Hinter Beerwalde senkt sich
die Strasse hinab ins romantische Thal der wilden Weisseritz, das hier
noch gar nicht durch Strassen aufgeschlossen ist und dem sicher eine
touristische Zukunft bevorsteht (Beerwalder Mühle Gasthof). Eine
directe Strasse führt aus dem Thal empor nach Oberröthenbach, eine
andere, etwas längere, durchläuft das ganze Dorf. Der
Oberröthenbacher Thurmberg (602 m) bleibt links und bald zeigt sich
in der Ferne die malerische Ruine von Frauenstein. Vom
Hartmannsdorfer Gasthof ab führt ein naher Fussweg (3½ km) nach
Frauenstein. Die Strasse berührt Kleinbobritzsch, wo der berühmte
Orgelbauer Gottfried Silbermann 1683 geboren wurde. Sein
Geburtshaus schmückt eine Marmortafel. Silbermann erlernte seine
Kunst in Strassburg bei einem Onkel und er brachte sie zu hoher
Vollendung. In Sachsen allein sind über 30 Orgeln von seiner Hand, die
berühmtesten stehen in der kath. Hofkirche zu Dresden und im Dom zu
Freiberg.
 Frauenstein. Rohland's Gasthaus zum Strauss. Gold. Stern. Gold.
Löwe. Das Parkschlösschen in reizender Lage an der Ruine ist z. Z.
ohne Wirth. 660 m. 1453 Einw. Das freundliche Städtchen wurde
nach dem Brande am 3. Oct. 1869 fast neuerbaut. Die
Frauensteiner Schlossruine ist zweifellos die imposanteste Ruine
des Erzgebirges auf Sächsischem Gebiet, sie lehnt sich dicht an das
neuere Schloss an und ihre isolirte Lage wie ihr Umfang machen sie
zu einer touristischen Notabilität. (Man meldet sich beim
Schlosshausmann.) Das interessanteste aus ihrer Geschichte ist eine
Belagerung durch Friedrich den Sanftmüthigen, der Heinrich II. von
Plauen seiner böhmischen Sympathien wegen daraus vertrieb. Das
neuere Schloss erbaute ein Heinrich von Schönberg. Vom Hause
Schönberg fiel die Burg mit ausgedehnter Herrschaft 1647 dauernd
an das Kurhaus. Den höchsten guterhaltenen Thurm nennt man die
Lärmstange, er wurde im 7jähr. Kriege von den Preussen zu
Lärmsignalen benutzt. Von seinen Zinnen hat man eine grossartige
Rundsicht; sie erstreckt sich gegen Norden bis ins Tiefland, davor
liegen die Thäler der Mulde, Bobritzsch und Weisseritz, sowie der
Tharandter Wald. Gegen Osten tauchen die Berge der Lausitz hinter
den Tafelbergen der Sächs. Schweiz auf, davor liegt die Dresdner
Gegend mit dem Borsberg, weiter herauf zeigen sich die Berge um
Altenberg und näher heran die Massenerhebung der Tellkuppe mit
ihrem Schwarzwald. Im Süden streicht der finster bewaldete Kamm
gegen die höchsten Berge des Gebirges, den Hassberg, Fichtel- und
Keilberg hin. Direct im Süden, der höchste Bergrücken mit einer
grossen Waldblösse trägt das einsame Lichtenwaldsteiner
Jagdschloss, seine Gipfel nennen sich Wieselstein und Schwarzeberg.
Im Nordwest schliessen die Augustusburg und der Rochlitzer Berg
den Umschaukreis ab. In der Nähe interessiren die imposanten
Trümmer. Der »dicke Märten«, etwas abgelegen, diente als
Burgverliess, doch ist er keineswegs ein »dicker«, sondern eher ein
schlanker Thurm. An einer Thurmruine sind 4 grosse Steinkugeln
eingemauert, wie sie ehedem zu Vertheidigungszwecken verwendet
wurden. Schöne Anlagen zu Füssen der Burg.
 Der Sandberg im Süden der Stadt mit Anlagen auf dem Gipfel
gewährt dieselbe Ausschau wie die Lärmstange. Empfehlenswerth ist
auch ein Spaziergang auf der aussichtsreichen Strasse nach Freiberg
bis zum Buttertöpfchen und dem Weissen Stein (1½ km). Es
sind das isolirte Quarzfelsen von seltsamer Gestaltung.
Die Wanderung geht nun bald durch Thalzüge oder auch über
aussichtsreiche Höhen nach Obernassau, Rechenberg (Liebschers
Gasthaus) und in die Rechenberger Waldregion durch den Fischerwald
nach böhm. Georgensdorf. Der sächs. Antheil mit grossem Gasthaus
heisst Georgenthal und hat nur 21 Einw. Auf böhm. Seite Lohse's
Gasthaus. Schön an der Flöha im weiten Thalkessel gelegenes Dorf.
Sommerfrische. 600 m. Herrliche Wälder. Bewegtes Terrain.

Anmerkung. Die Wanderung nach dem Lichtenwaldsteiner

Jagdschloss und die unendlich lange Waldschneusse nach dem
Wieselstein und der in der Nähe gelegenen Försterei Georgenhöhe
ist landschaftlich nicht sonderlich lohnend; auch ist der Wieselstein
völlig verwachsen, nur der rechts abgelegene Schwarzeberg hat
seine grossartige Aussicht auf Böhmen noch nicht eingebüsst, doch
ist derselbe ohne Führer schwer zu finden. Das Lichtenwaldsteiner
Jagdschloss mit seiner Geweihsammlung hat in seiner sehr einsamen
Lage allerdings touristischen Reiz. Ueber die Höhe der beiden Gipfel
des langen Bergrückens ist etwas Sicheres nicht zu erfahren
gewesen. Georgenhöhe, die Försterei, liegt 871 m hoch, doch ist
auch diese Zahl unter Reserve gegeben. Der Rücken dürfte 930 m
nahekommen. Wer diese Partie unternimmt, sorge für Proviant.
Der Tour treu gehen wir im schönen Waldthal an der Flöha
aufwärts. Gleich im Anfang links imposante Felspartien. Fley liegt
unfern des Ursprungs der Flöha. Das Kirchdorf scheint seinen Namen
dem Fluss entliehen zu haben. Im Winter Nachm. 3 Uhr
Wildfütterung, zu der sich circa 350 Stück Hochwild einstellen. Die
Strasse führt nun durch einsamen Hochwald nach
Langenwiese. Langes Dorf, frei auf dem Kamm gelegen mit einer
herrlichen Aussicht auf Eger- und Bielathal, wie auf das böhm.
Mittelgebirge. Der Wieselstein, wenn ein Aussichtsthurm errichtet
wäre, würde freilich eine volle Rundschau darbieten, er beherrscht
die ganze nördl. Abdachung bis Augustusburg hin und den Kamm
des Gebirgs selbst. Vom Erzgebirge selbst sind die höchsten
Erhebungen Keil- und Fichtelberg sichtbar. In der Ferne taucht das
Karlsbader- und dahinter das fränkische Fichtelgebirge auf. Zu
Füssen dominirt besonders das imposante Schloss Eisenberg.
Auf steilem Pfad steigen wir in das Ossegger Thal, an dessen
Gehängen die Ruinen der Riesenburg sich erheben und an dessen
Mündung das stolze Ossegger Kloster sichtbar wird. Die
Riesenburg, welche im 14. Jahrh. von den Hussiten zerstört wurde,
ist die Geburtsstätte des Apostels der Preussen, des hl. Adalbert.
Vom Wartthurm herab überaus anmuthige Thalbilder sowie
Fernblicke auf das Mittelgebirge und den Dux-Teplitzer Thalkessel.
Ossegg und weiter nach Teplitz siehe Routennetz.

8. Dresden-Tharandt (15 km). Klingenberg

(11 km). Freiberg (14 km).
Man fährt wie bei Tour 7 durch den industriereichen Plauenschen Grund
nach Tharandt. (Siehe Seite 41.) Von hier steigt die Bahn 1 zu 40 an
der Weisseritz, später am Serrenbach aufwärts nach Klingenberg (432
m) und auf die Freiberger Hochebene. Bei Passirung der Freiberger
Mulde links die grossartigen Muldenhütten, fiscalische
Silberschmelzwerke. Die hohe Esse auf der Anhöhe ohne Gebäude
daneben ist die Giftesse, welche die Arsendämpfe ableitet. Wer die
Hütten besuchen will, steigt am Besten hier aus und geht dann die 3½
km nach Freiberg zu Fuss. (Muldenhütten siehe Seite 48.)

Freiberg. Hotel de Saxe, Hirsch, Stern, Adler, Goldne Pforte,

Preuss. Hof, Stadt Altenburg. Zu besuchtesten Wirthschaften gehört
Debus, Otto, Reichel. Gartenwirthschaften: Schillerschlösschen,
Schützengarten, Union, Feldschlösschen, Tivoli. Bäder: Actienbad,
Stadtbad, Hedrichs Bad. Droschken: Einfache Fahrt 50 Pf.
Freiberg am Münzbach, unfern der Freiberger Mulde, liegt 414 m
hoch in einer Gegend, wo die Bodenplastik des Erzgebirges nicht
sehr energisch ist, doch gehört die Stadt selbst zu den
sehenswerthetesten in deutschen Landen; sie ist das Centrum des
erzgeb. Bergbaus, der nicht nur die starke Besiedlung Sachsens
verursacht, sondern den Grundstock der eminenten
Gewerbthätigkeit des ganzen Landes bildet. Freiberg ist noch heute
der Stolz des Sächs. Fürstenhauses, wie der des ganzen Landes.
25 431 Einw.
Geschichtliches. Wenn man den vielen slav. Worten, die noch heut
im Sächs. Bergbau gebräuchlich sind, historische Bedeutung
beilegen darf, so haben die Sorben bereits Bergbau getrieben und
wohl auch die Schätze der Freiberger Gegend, die ja zum Theil von
Slaven besiedelt war, gekannt, doch der Aufschwung, der Betrieb im
grossen Styl ist den germanischen Elementen zu verdanken.
Slavische Bezeichnungen sind: Zscherper, Zeche, Stollen, Schwaden,
Rösche, Kux, Kobalt, Flötz, Druse u. a. m.
Die Sage berichtet, Fuhrleute aus Halle oder Goslar, die Salz und
Blei geladen hatten, fanden im Fahrgleis eine Erzstufe, die sie in
Goslar untersuchen liessen. (Siehe auch Geschichtliches.) Darauf hin
seien denn Harzer Bergleute in Schaaren in das rauhe Waldgebirg
gezogen. Thatsache ist, dass der Freiberger Stadttheil, die
Sächsstadt, von niedersächsischen Bergleuten ihren Namen erhielt.
Aber zur Besiedlung hat wohl ganz Deutschland und vor Allem das
nahe Böhmen beigesteuert. Otto der Reiche, der seinen Beinamen
den Freiberger Erzreichthum verdankt, erhob die Ansiedlung 1175 zu
einer Bergstadt mit vielen Vorrechten, wie auch der Name Freiberg
darthut. Der Glanz und der Ruhm der jungen Ansiedlung machte den
Hohenstaufenkaiser Heinrich VI. lüstern auf die Stadt; schwäbisches
Kriegsvolk besetzte sie, dieses wurde aber von der, dem Landesherrn
treu ergebenen Bürgerschaft vertrieben. Dem Kaiser Adolph von
Nassau gelang es wirklich, den Landesfürsten, Friedrich mit der
gebissenen Wange, zu vertreiben, doch nach wenigen Jahren
schüttelten die Bürger die kaiserliche Herrschaft für immer ab. Im
Anfang des 16. Jahrh. gelangte Freiberg auf den Zenith seines
Ruhmes, die Stadt zählte damals 70 000 Einw., war also eine der
volkreichsten Städte deutscher Zunge in damaliger Zeit. Während
des 30jähr. Krieges sank sie zu einem Schatten ihrer ehemaligen
Grösse herab. Zwar belagerte Banner und Torstensohn die Mauern
vergeblich, aber die meisten Häuser gingen in Flammen auf, die
Vorstädte lagen verwüstet, die Bergwerke ausgeraubt und verödet
oder gar ersoffen, die Bevölkerung war bis auf 6000
zusammengeschmolzen. Im 7jähr. Kriege wurde das Gemeinwesen,
das immer als reich galt, mehrfach gebrandschatzt und Prinz
Heinrich, der Bruder Friedrich II., schlug eine bedeutende Schlacht
bei Freiberg zu Friedrichs Gunsten. Auch in den Franzosenkriegen litt
die Bergstadt durch Einquartirung.
War es früher der Hofhalt der reichen meissnischen Fürsten, der
ausser den metallischen Schätzen, Glanz und Ansehen über die Stadt
gebreitet, so waren es später hochberühmte Männer, die hier gelebt
und gewirkt. Genannt seien Werner, der grosse Geologe, Alexander
von Humboldt, Lampadius, der Erfinder des Leuchtgases, der
Geologe Breithaupt, Silbermann, der Orgelbauer, Theodor Körner als
Bergstudent und der jugendliche Carl Maria von Weber. In den
Wissenschaften des Bergbaus und der Hüttenkunde steht Freibergs
Name noch immer obenan, im Bergbau selber hat es die Priorität an
die colossalen Minen Nevadas abtreten müssen. Der Grubenbau
erweist sich zwar ergiebiger denn je, allein es ist nicht der
Erzreichthum die Ursache, es ist das ein Triumph der Hüttenkunde,
die jetzt die ärmsten Erze mit besseren Hilfsmitteln auszunützen
versteht.
Neben dem Bergbau blüht auch eine bedeutende Industrie in
Freiberg, so die Superphosphatfabrikation, die sich auf den Bergbau
stützt, ferner bestehen Cigarrenfabriken, eine grosse
Portefeuillefabrik, eine Flachsspinnerei und eine Goldtressenfabrik,
die für die deutsche wie für fremde Armeen die Tressen liefert.
Bergmännisches. Zur völligen Kenntniss der Bergstadt gehört eine
Einfahrt in einen der Schächte, die zwar nicht mühelos, aber
hochinteressant ist. Erlaubnisskarte auf Himmelfahrt 2 Mk., 2
Personen 3½ Mk. Auch Damen fahren zuweilen an, doch müssen
auch sie ein männliches Berghabit anlegen. Grösste Tiefe 520 m.
Man denke ja nicht, dass man in den engen feuchten Stollen
glänzende Silberadern zu sehen bekommt. Das gediegene Erz ist
sehr selten. Auf 200 Ctr. gesprengte Gangmassen kommt erst 1 Ctr.
Erz und dieses führt im Durchschnitt nur 1 bis 2% Silber bei sich.
(Die Silberbergwerke sind weit gefahrloser wie die
Kohlenbergwerke.)
Der Bergmannsstand hat durch seine genossenschaftliche
Organisation, seinen conservativen Sinn und seinen Berufsgeist
vieles aus alter Zeit herübergerettet, so dass sich der Bergmann von
seinen »überirdischen« Standesgenossen, den Fabrikarbeitern
unterscheidet. Schon seine Sprache hat Eigenartiges, so sind Kux,
Zeche, Zubusse, Wetter, Hund, Teufe etc. ihm ureigene Begriffe. Als
Scheidejunge tritt der jugendliche Bergmann seinen Beruf an, er
sortirt zunächst die ausgebrachten Erze nach ihren Gehalt, sodann
steigt er zum Grubenjungen empor oder vielmehr, er steigt hinab
und vermittelt den Verkehr zwischen den Abbauörtern und den
Förderschächten, indem er die Hunde (Fahrzeuge) mit den Erzen
hin- und hertreibt, oder er räumt Gesteinstrümmer bei Seite. Als
Lehrhäuer hat er sieben lange Jahre zu dienen, ehe er zum
Probegeding zugelassen wird; ist dieses gemacht, dann erst gilt er
als ein vollbürtiger Bergknappe. Eine ähnliche Stufenleiter haben die
Zimmerlinge zu durchlaufen. Obersteiger müssen eine Bergschule
absolvirt haben. Die Kleidung besteht in einem Grubenkittel, einem
Bergleder und einem Berghut ohne Krämpe mit einer Kokarde von
gekreuzten Fäusteln gebildet. Bergaufzüge mit Fackeln und
Grubenlichtern sind leider seltener geworden. Der grösste neuere
Bergaufzug dürfte der 1878 zu Dresden gewesen sein, der zu Ehren
der Silbernen Hochzeit des Königspaares stattfand. Die hohen
Bergbeamten tragen glänzende Uniformen. Alt wird der Bergmann
selten, nahe den Fünfzigen siegt er an der sogenannten Bergkatze,
einer Art Schwindsucht dahin, den Hüttenleuten ergehts fast
schlimmer, sie erliegen meist noch früher der Hüttenkatze, einer
ähnlichen Krankheit. Am gefährlichsten ist die Arbeit in den
Arsenhütten. Die Silberbrenner leiden häufig an der Bleikolik. Die
reichste Grube ist gegenwärtig Himmelfahrt, sodann folgt
Himmelsfürst. Neuerdings werden auch von mehreren andern
Schächten reiche Anbrüche gemeldet.
Die Muldenhütten (3½ km) an der Mulde sind hochintressant.
Eintritt 1 Mk. Die gepochten Erze werden zunächst in Röstöfen
geröstet, d. h. vom Schwefelgehalt befreit, sodann geschmolzen,
wobei sich das Werkblei ausscheidet. Dieses geht nun durch viele
Oefen, ehe es zum Silberbrenner gelangt, der mit Feuer und
glühenden Sauerstoffströmen den Bleigehalt zum Oxidiren zwingt,
bis der Rückstand den sogenannten Silberblick zeigt und damit sich
als reines Silber ausweisst. (Näheres siehe des Verfassers Aufsatz
Gartenlaube 1879 No. 34.) Nebenher sind Bleirohrfabriken,
Arsenhütten, Schwefelsäurefabriken, Zinköfen,
Goldscheidewerkstätten zu besichtigen. Auch 2 kleine Platinkessel
sind intressant, sie kosten zusammen 36 000 Thlr.
Anmerkung. Wer die Grabentour machen will, besucht anstatt der
Muldenhütten die am Wege gelegenen Halsbrücker Hütten, die ganz
ähnlich eingerichtet sind.
Die Bergacademie geniesst Weltruf. 15 acad. Lehrer unterrichten
gegen 150 Bergstudenten unter denen alle Welttheile vertreten sind.
Grossartige Mineraliensammlung. Fachbibliothek von circa 40 000
Bänden. Alexander von Humboldt und Theodor Körner studirten hier.
(Neben der Bergacademie sei das Gymnasium genannt, das schon
seit 4 Jahrh. besteht und neuerdings in einen stattlichen Neubau
übergesiedelt ist.) Alterthumsmuseum im Kaufhaus. Sonn- und
Feiertags von früh 11 bis 12 Uhr und Nachmittags von 2 bis 5 Uhr,
10 Pf. Eintritt, Mittwoch und Sonnabend Nachmittag 20 Pf. Ausser
der Zeit Führung 1 bis 4 Personen 1 Mk. Katalog 20 Pf. 1861
gegründet. Von localem Interesse sind besonders alte Berggeräthe,
die man in verlassenen Bauen aufgefunden. Ferner kirchliche
Alterthümer, Innungsladen, Trinkgefässe etc.
Interessante Gebäude sind das alte Rathhaus, in dem auch ein
Stück der Strickleiter aufbewahrt wird, auf welcher Kunz von
Kauffungen den Prinzenraub ausführte, ferner das Kaufhaus, das alte
Gymnasium, sonst Domherrenwohnung, und viele hochgegiebelte
alte Bürgerhäuser.
Der Freiberger Dom ist weniger berühmt geworden durch seinen
Gesammteindruck, wie durch seine Goldene Pforte und seine
Grabmäler. 1484 abgebrannt, wurde er bald darauf in spätgoth.
Styl wieder errichtet, doch hat man den kostbarsten Ueberrest von
dem alten romanischen Bau wohlerhalten. Diese goldene Pforte ist
von den älteren Werken der Steinplastik zweifellos das
hervorragendste in sächs. Landen und überhaupt einzig in ihrer Art.
Figurenreiche Gruppen, darunter die Dreieinigkeit, die Anbetung der
heil. drei Könige, Madonna, Engel, Apostel, Abraham, Josua, David,
auferstehende Todte etc. beleben das geniale Bildwerk, dessen
Schöpfer eine hohe künstlerische Anmuth bei reicher Phantasie und
strenger Anordnung bekundet. Bemerkenswerth sind noch der
Kreuzgang in edler Gothik. Im Innern des Domes eine ältere Kanzel,
die eine Riesentulpe darstellt; im Kelch befindet sich der
Predigtstuhl. Die zweite Kanzel, gestützt durch zwei bergmännische
Statuen, liess der Freiberger Bürgermeister Schönleben 1638
errichten; man schätzt sie als eines der besten Werke altdeutscher
Renaissance. In der anstossenden Begräbnisscapelle liegen die
Fürsten Sachsens von Heinrich dem Frommen bis auf Johann Georg
IV. Das künstl. bedeutendste Grabmal ist das des Kurfürsten Moritz,
der bei Sievershausen gefallen. Die Zeichnung stammt von zwei
Italienern, den Gebrüdern Tola, die Ausführung in Marmor geschah
durch den Antwerpener Bildhauer van Zerun. Die Orgel zählt zu
Silbermanns Meisterwerken. Auf dem eingezogenen Domkirchhof ist
das Grab des grossen Geologen Werner erhalten worden.
Spaziergänge. In der Stadt besichtige man den blauen Stein auf
dem Obermarkt, der die Stelle bezeichnet, wo Kunz von Kauffungen,
der Prinzenräuber, seinen Putsch mit dem Leben büsste. An einer
Ecke des Rathhauses ist ein Kreuz von besonders reichen Silbererzen
eingemauert. Ein Rundgang um die in Promenaden verwandelten
Ringwälle ist höchst empfehlenswerth. Wir berühren dabei am
Kreuzthor das Denkmal des Mineralogen Werner, vor dem
Petersthor steht das renovirte Schwedendenkmal. An den
Bürgermeister Horn erinnert vor dem Erbischen Thor ein goth.
Brunnen. Zwischen dem Schloss Freudenstein, wo einst Heinrich
der Fromme fröhlichen Hof hielt, und den idyllischen Kreuzteichen
steht das Kriegerdenkmal auf einer alten Schanze.
Wer eine echt bergmännische Landschaft mit Zechenhäusern und
grossen Berghalten sehen will, gehe nach der Sächsstadt und in das
Schachtrevier von Himmelfahrt. Ein Besuch von Herders Ruhe, ein
Denkmal unfern der Strasse nach Halsbrücke kann damit verbunden
werden. Schöner Blick auf die alte Bergstadt, der in Deutschland
wohl einzig in seiner Art sein dürfte. Grössere beliebte Spaziergänge
sind die nach Fernesiehen, ein Gasthaus an der Chemnitzer
Strasse, nach dem Rosinenhäuschen an der Frauensteiner Strasse
und nach Heedens Rest. an der Meissner Strasse.
 Touren ab Freiberg.

9. Freiberg-Halsbrücke (5 km).
Krummhennersdorfer Mühle (4 km).
Oberreinsberg (4 km). Zollhaus (1½ km).
Bergwerk Kurprinz (über Burkersdorf 4 km).
Ueber Altväterwasserleitung nach Freiberg (8 km).
Wir wandern, Herders Ruhe und später Tuttendorf zur Rechten, hinab ins
Muldenthal nach Halsbrücke. Grossartige fiscalische Hüttenwerke,
Silberschmelzen, ganzähnlich eingerichtet, wie die Muldenhütten.
Führung 1 Mk. Bei Halsbrücke befinden sich auch grosse Bingen. Die
Einbrüche geschahen 1662 und 1691. Das Muldenthal wieder
verlassend, wandern wir über die Höhe, dann durch
Krummhennersdorf, wo einst Markgraf Albrecht der Stolze an Gift
starb, hinab nach der stattlichen Mühle an der Bobritzsch. Dieser Fluss
ist der stärkste Nebenfluss der Mulde und hat prächtige Thalpartien.
Bei der Mühle beginnt die Grabentour, so nennt sich der schöne
Prommenadenweg im romantischen Thalzug, der allen Krümmungen
des Berggrabens folgt. Hier stossen wir auch auf Lichtlöcher des
Rothschönberger Stollens, einer der grössten Bergstollen. Sein Bau
begann 1840, seine Vollendung fällt ins Jahr 1879. Die Länge beträgt
15 km, erreicht also die des St. Gotthardtunnels.
Am 6. Lichtloch vorüber, verlässt der Pfad bald das Thalgehäng und führt
nach Oberreinsberg hinauf, dessen Schloss mit Kirche schon lange
sichtbar ist. Nun hinab an das Zollhaus, ein romantisch gelegenes
Gartenrestaurant unfern des Einflusses der Bobritzsch in die Mulde
(von hier nach Nossen im schönen Thalzug Über die Steier- und die
Beiermühle 6 km. Nossen s. unten.). Der Rückweg führt uns zunächst
nach Bieberstein, schöner Schlosspark mit Ruine der alten Burg
Bieberstein. Ueber Burkersdorf und Teichhäuser gelangen wir wieder
hinab ins Muldengebiet und zum Kurprinzen, einem der beliebtesten
Ausflugsorte der Freiberger. Im Huthaus Schenke. Schöne
Gartenanlagen vom Oberberghauptmann Frhrn. von Herder
 herrührend. Die Altväterwasserleitung, die einst einen
Bergwerkscanal über das Muldenthal leitete, liegt etwas abseit des
Weges nach Freiberg. Die ganze Anlage erinnert an römische
Aquaducte und in einiger Entfernung giebt sie ein imposantes, bei uns
seltenes Ruinenbild.

10. Freiberg-Nossen-Altzella. Bahnausflug

(25 km).
Die Bahn durchschneidet den Freiberger Spitalwald und gelangt später in
den grossen Zellaer Waldcomplex.

Nossen (slav. Nozzin). Stadt Dresden. Blauer Stern. Deutsches

Haus. 258 m. 3700 Einw. Anmuthig auf einer Höhe über der Mulde
gelegenes sauberes Städtchen. Das besonders vom Thal aus
imposante Schloss ist gegenwärtig Sitz der Gerichtsbehörden. Den
ältesten Theil, die sogenannte Dechantei, hat das Zwickauer
Arbeitshaus in Beschlag genommen als Zellengefängniss. Beim
Neubau der Nossener Stadtkirche 1563 erhielten die Bürger die
landesherrliche Erlaubniss, von den Ruinen des nahen Klosters Zella
Mauertheile abtragen und verwenden zu dürfen, daher die
romanischen Seitenportale in dem sonst nüchternen Bau. Im südl.
Portal ist die sogenannte Rose von grösserem kunsthistorischen
Werth. Die darüber hängende Riesenrippe soll einer lustigen Sage
nach einem Fräulein von Neudeck zugehören.
Altzella (2 km von Nossen). Ausgedehnte, freilich schwer
beraubte Ruinenstätte des 1545 säcularisirten Cistercienserklosters
Altenzella. Ehedem markgräfl. Begräbnissstätte und mächtigstes
Kloster des Meissn. Markgrafenthums. Begraben liegen hier nach
den Inschriften: Otto der Reiche, Albrecht der Stolze, Dietrich der
Bedrängte, Heinrich der Erlauchte, Friedrich der Ernsthafte und
Friedrich der Strenge und die Gemahlinnen dieser Fürsten. Die
Fürstengruft, 1787 restaurirt, erfreut sich einer gewaltigen Akustik,
die der Führer durch Gesänge weckt. Auf einigen Grabsteinen finden
sich die Bildnisse mehrerer Fürsten in ziemlich kunstloser Weise
dargestellt. Im Park liegen verstreut viele Ueberreste des alten
Klosters und seiner Begräbnisskapellen, die von alten meissnischen
Adelsgeschlechtern benutzt wurden. In der Umfassungsmauer gegen
Westen ein Thor in romanischem Styl. In den Wirthschaftsgebäuden
des jetzigen Kammergutes finden sich gleichfalls noch viele
Ueberreste des einst so mächtigen Klosters. Wie belebt die Stätte
war, das möge eine Notiz aus dem Anfange des 15. Jahrhunderts
illustriren, wonach im Jahre durchschnittlich 6000 Fremde zu Fuss
und 5000 zu Pferd im Kloster einsprachen.

11. Freiberg-Mulda (15 km). Bienenmühle

(12 km). Georgensdorf (5½ km). Fley (7 km).
Langewiese (5½ km). Ossegg (über Riesenburg
5 km). Teplitz (10 km).
Bis Bienenmühle ist Eisenbahnfahrt zu empfehlen, doch ist der Weg im
Muldenthal aufwärts oder über Lichtenberg keineswegs ohne
Anmuth. (Das gilt auch von der directen Strasse nach Frauenstein.
Bequemer erreicht man freilich die inter. Stadt mit ihrer Ruine von
Station Mulda aus. 10 km. Man geht am 615 m hohen Burgberg
vorüber und berührt kurz vor Frauenstein den Weissen Stein und das
Buttertöpfchen, intr. Quarzfelsen. Frauenstein s. S. 43.)
Hinter Mulda wird das Muldenthal einsamer und romantischer. Bei
Bienenmühle findet die Eisenbahn ihre vorläufige Endstation.
Gasthaus zu Bienenmühle gut und billig. Anmuthige Thallage des im
Entstehen begriffenen Ortes. Ein schöner Fussweg geht von hier nach
Neuclaussnitz und an der sogenannten Säueck vorüber nach
Georgenthal und Georgensdorf. Die Strasse, Klötzerweg genannt,
führt gleichfalls durch herrliche Waldungen. Auf der Höhe beim
Wegweiser nach Cämmerswalde einige Hundert Schritt links gehen, wo
sich ein Wegweiser nach Georgenthal findet. Georgenthal-Georgensdorf
und weiter siehe Routennetz.
 12. Freiberg-Bienenmühle (15 km).
Cämmerswalde (5 km). Purschenstein (5 km).
Bad Einsiedel (5 km). Böhmisch Einsiedel
(2 km). Dorf Kreuzweg (5½ km). Oberleitersdorf
(6 km). Ossegg (8 km). Teplitz (10 km).
Von Freiberg bis Bienenmühle s. Tour 11. Von hier gleichfalls den
Fussweg über Neuclaussnitz und die Säueck oder die Strasse
(Klötzerweg) nach Cämmerswalde und an der Flöha hinab nach
Purschenstein.

Purschenstein. Auf einem Felsen an der Flöha gelegenes

Schloss, der Familie von Schönberg gehörig, war ehemals ein
böhmisches Krongut. Ahnensaal mit Bildnissen Schönberg'scher
Familienmitgliedern. Kostbare Uhr, mehrfach prämiirt. Im Winterhaus
tropische Nadelhölzer. Im Park ein goth. Thorhaus und eine hölzerne
Einsiedelei, unter welcher sich die Familiengruft eingemauert
befindet. Purschenstein mit seinen finsteren Thürmen ist eines der
imposantesten Schlösser des Erzgebirges. Das nahe Neuhausen mit
schöner, neuerbauter gothischer Kirche gehört zu den sieben
Spielwaarendörfern des Seiffener Bezirks (siehe unter Seiffen).
Erbgericht, gutes Landgasthaus.
In Windungen führt die Strasse aufwärts nach Bad Einsiedel. (Das Bad
steht etwas links der Strasse, die auf der Höhe schöne Ausblicke auf
den Olbernhauer Grund und die benachbarten Wälder gewährt.) Bei
einem Strassenknie geht ein angenehmer Fussweg durch den Wald an
das Bad.

Bad Einsiedel. Zur Herrschaft Purschenstein gehörig, vom

Forstwirth Ueberschaar bewirthet. 751 m hoch gelegen.
Schwefelhaltiger Eisenquell. Gegen 200 Badegäste. Beliebte
Sommerfrische. Viel Passanten. Ausflüge nach dem Schwartenberg
(s. unter Seiffen) und dem 833 m hohen Ahornberg (nach Böhmen
beschränkt). Herrlich ist die Aussicht bei Göhren (4 km von Bad
Einsiedel entfernt) auf Böhmen hinab. Touristen gehen dann von
Göhren direct den Rauschengrund hinab nach Oberleitersdorf.
 In Dorf Einsiedel auf böhm. Seite das grosse Dietel'sche Gasthaus. Die
Strasse windet sich auf den Kamm hinauf und fällt dann rasch ab nach
Dorf Kreuzweg. Gasth. zur Waldburg. Herrliche Fernsicht auf das
eisenbahndurchzogene Eger- und Bielathal und auf das böhm.
Mittelgebirge. Die schönste Aussicht bietet die Waldblösse unfern der
Waldburg. Ein directer Weg führt durch den Hammergrund nach
Hammer. Johnsdorf ist Bahnstat. der Kommotau-Dux-Bodenbacher
Bahn. In Hammer herrlich gelegenes Rest. zur »Deutschen
Bruderhalle« mit ebenso schattigem als aussichtsreichem Garten.
Hammer eignet sich zur Sommerfrische.

Oberleitersdorf. Drei Linden. 5300 Einw. Industrielles Städtchen,

das Centrum der böhm. Spielwaarenindustrie. Vom Schiesshaus
schöne Ausblicke auf Thalbecken und Mittelgebirge.
Von hier mit Dampf oder die aussichtsreiche Strasse über Ladung nach
Ossegg und Teplitz. S. S. 40 und 21.

13. Freiberg-Sayda (mit Bahn bis Nassau 21,

von hier über Claussnitz 10½ km, oder die Strasse
über Brand und Grosshartsmannsdorf 28 km).
Purschenstein (5 km). Bad Einsiedel und weiter
nach Oberleitensdorf, Ossegg, Teplitz. S. Routennetz.
Nach der ersten Variante geht man von Stat. Nassau nach Claussnitz und
über Friedebach nach Sayda. Wer auf Sayda verzichtet, geht von
Claussnitz direct über den 730 m hohen Meiseberg nach Purschenstein
(12 km). Im anderen Falle wandert man von Freiberg durch das
dichteste Schachtrevier nach Brand (6 km). Am Wege viel
bergmännisches Leben und links und rechts ertönen Bergglöckchen.
Brand. Zum Kronprinz. Goldner Stern. Rest. Rathskeller. Bergstädtchen,
2818 Einw., die einzige Stadt Sachsens ohne Kirche (ist nach Erbisdorf
eingepfarrt). Von hier nach Grosshartmannsdorf (8 km). Langes
Dorf, an dessen Ende ein 60 hectaren grosser Bergteich liegt. 1880
entnahm man demselben 380 Ctr. Fische. Weiter führt uns der Weg an
dem 711 m hohen Saydenberg vorüber, ein Berg mit breitem, flachem
Gipfel. Die Ersteigung ist mühelos und verschafft einen umfassenden
Blick hinauf in das Centralerzgebirge und auf die Olbernhauer Gegend.
 Bald berühren wir Dörnthal. 3½ km von der Strasse abseits liegt der
seeartige Dörnthaler Bergteich, von welchem aus der 28 km lange
Dörnthaler Kunstgraben beginnt, der dem Freiberger Bergbau dient.
Unfern des Teiches liegt die Mündung des ehemals berühmten
Friedrich-Bennostollens, der für Kähne schiffbar und so hoch ist,
dass ein Reiter passiren könnte. Früher wurde der noch heute
imposante Bau öfter illuminirt, wenn fürstl. Personen anwesend waren.

Sayda. Löwe. Stern. Ross. Rest. zum Rathskeller. 1612 Einw. 677
m. ü. M. Das freundliche Städtchen ist nach dem Brande 1842 fast
neuerbaut. Sayda war im frühesten Mittelalter eine wichtige
Handelsetappe zwischen Böhmen und dem Norden Deutschlands.
Eine Judenstadt ist verschwunden, auch von der alten Sorbenburg
Saydowa ist kein Stein mehr zu sehen. In der Kirche Grabmäler vom
Bildhauer Nosseni, der Familie von Schönberg zugehörig. Von der
Thurmgallerie grosse Umschau über die Olbernhauer Gegend und
über das Centralerzgebirge bis zur Augustusburg.
Nach Purschenstein hinab geht man den angenehmeren Fussweg am
Wald, die Strasse zur Linken lassend. Purschenstein und weiter siehe
Routennetz.

14. Freiberg-Oederan (17 km). Flöha (10 km).

Chemnitz (12½ km).
Die Bahn überwindet mit geringen Curven die kleinen Terrainfalten der
Freiberger Hochebene, bis sie vor Oederan in eine bewegtere Gegend
gelangt, deren Mittelpunkt die stattliche Augustusburg bildet.

Oederan. Hirsch. Deutsches Haus. Bellevue. Garküche. Rest.:

Rathskeller. Kögel. Günther. 5850 Einw. 383 m ü. M. Flanell- und
Tuchfabriken. Anmuthige Lage. Nur 3 km entfernt über das
hochgelegene Rittergut Börnichen liegt die 482 m hohe
Schönerstädter Höhe, welche ihrer prächtigen Aussicht wegen
neuerdings viel besucht wird.
Die Bahn windet sich hinab an die Gehänge des Flöhathals, überschreitet
bei Hetzdorf auf dem imposanten Hetzdorfer Viaduct die Flöha und
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookultra.com