Bioinformatics Advice on Experimental Design

Where do I start?
Please refer to the following guide to better plan your experiments for good statistical
analysis, best suited for your research needs. Statistics cannot rescue a bad experimental
design.

Please contact our Bioinformatics team for a consultation when in doubt.

Next Generation Sequencing (NGS) experiments

Many steps in the experimental process can introduce various biases and errors, and careful
consideration must be given to the following aspects:

§ Platform choice:

Platform Platform
Genome Sequencer Genome Hiseq 2000 SOLiD 4 system HeliScope
FLX Titanium System Analyzer IIx
Company Roche Illumina Illumina Applied Biosystems Helicos
Biosciences
Read length 400-600bp 2x100bp 2x100-150bp 50 +25bp ~30bp
Samples per run 16 8 16 16 50
Reads per run ~1 million ~300million ~800 million >700 million ~500 million
Run time 10 h 8 days 8 days 11-13 days 8 days
Website www.454.com www.illumina.com www.illumina.com www.appliedbiosystems.com www.helicosbio.com

These numbers change rapidly as technology improves. Please note that these numbers are
based on data from Oct. 2010. Please refer to the websites listed under each platform for the
latest numbers.

§ Type of Run – Paired End (PE) or Single End (SE):

The following table provides a guide to what type to run is recommended for typical
applications of various NGS assays.

Center for Research Informatics Bioinformatics Core last updated May 2015
 Paired End Single End
RNASeq - De novo Assembly RNASeq - Counting
RNASeq - Splicing ChIP-Seq - Counting
ChIP Seq – Epigenetic modifications
DNA – SNP Identification
DNA – Indel identification
DNA – Structural variants

§ Read Length:
50bp reads are typically sufficient for read mapping to the reference genome, and
RNASeq counting experiments. >100bp reads are useful for whole genome and
transcriptome studies based on the application.

§ Replication:
Samples must be sequenced with replicates to identify sources of variance and increase
statistical power to separate true biological variance from technical variance. Biological
replicates are critical whereas technical replicates are typically not required.

Cutting back replicates to reduce cost might seem like a good option, but remember: A sample or
sequencing run can fail, and lead to repeating the experiment.

In general, 4 biological replicates per experiment are recommended, however, 3

replicates if also reasonable. Please consult with us with further questions. You can also
use http://bioinformatics.bc.edu/marthlab/scotty/help.html for calculation of power
from your pilot data.

§ Randomization:
Assign individuals at random to different groups to reduce bias. We recommend
randomization of samples such that each sequencing lane contains samples from all
experimental groups. Please refer to Blocking and Multiplexing below to understand how to
do this.

§ Blocking & Multiplexing:

Distribute samples across various lanes on the flowcell to avoid lane effects. Use
multiplexing effectively for balanced block designs. (Fig.1) But all samples cannot be
sequenced on each lane as the number of unique barcodes for each lane also limits us.
Solution: Balanced incomplete block design.

“Block what you can and randomize what you cannot.” – Box, Hunter, & Hunter (1978)

Center for Research Informatics Bioinformatics Core last updated May 2015
 Group! A! B!

Biological 1! 2! 3! 1! 2! 3!
replicates!

RNA
R1! R2! R3! R1! R2! R3!
extraction!

Flowcell! Flowcell!

L1! L1! L1! L1! L1! L1!

L2! L2! L2! L2! L2! L2!
L3! L3! L3! L3! L3! L3!
L1! L2! L3! L4! L5! L6! L4!
L5!
L4!
L5!
L4!
L5!
L4!
L5!
L4!
L5!
L4!
L5!
L6! L6! L6! L6! L6! L6!

Lane1 Lane2 Lane3 Lane4 Lane5 Lane6! Lane1 Lane2 Lane3 Lane4 Lane5 Lane6!

✗ ✔
If,
I= Number of groups/treatments
J= Number of biological replicates per treatment
s= Number of unique barcodes that can be added in one lane
L= Number of lanes sequenced
T=Total number of technical replicates
sL
T=
JI
If s<I, complete block design is not possible. [1]

§ Sequencing depth:
The following table provides general recommendations for coverage/reads
(https://genohub.com/recommended-sequencing-coverage-by-application/) for typical
read lengths for the HUMAN genome. Please visit https://genohub.com/next-
generation-sequencing-guide/#reads for typical number of reads/lane for various
commonly used NGS platforms.

A useful resource from Illumina for specific coverage estimates for various Illumina
instruments and genomes of different sizes is
http://support.illumina.com/downloads/sequencing_coverage_calculator.html

Center for Research Informatics Bioinformatics Core last updated May 2015
 DNA:

Category Application Recommended coverage

(X) or reads (in millions)
Whole genome Re-sequencing 30-80X
De novo assembly 100X
SNP detection 10-30X
Indel detection 60X
Genotype calls 35X
CNVs 1-8X
Whole exome SNVs detection 100x (3-13x local depth)
sequencing
Indel detection NOT recommended
De novo assembly >100M
DNA Target- ChIP-Seq 10-40X [10-14M (sharp
Based peaks); 20-40M (broad marks)]
Sequencing
Hi-C 100M
DNA CAP-Seq >20M
Methylation
Sequencing
RRBS (Reduced Representation 10X
Bisulfite Sequencing)
Bisulfite-Seq 5-15X; 30X

RNA (for human/mouse genome):

Please note that the number of reads your need for any type of RNASeq also depends on
the desired dynamic range of expression.

Category Application Recommended of mapped

reads (in millions)
Transcriptome Differential expression 10-25M
Sequencing
(RNA-Seq)
Alternative splicing 50-100M
Allele specific expression 50-100M
RNA-Target- CLIP-Seq 10-40M
Based
Sequencing
PAR-CLIP 5-15M
RIP-Seq 5-20M
Small RNA Differential Expression ~1-2M
(microRNA)
Sequencing
Discovery ~5-8M

Center for Research Informatics Bioinformatics Core last updated May 2015
 Microarray Experiments
A very useful resource for microarray design is:
http://discover.nci.nih.gov/microarrayAnalysis/Experimental.Design.jsp
• Balanced samples
o Same amount of cases and controls
o Matched phenotypes: gender, age, etc.
• Biological replicates
o Pure background to avoid biological variation
o More replicates are needed if there is larger variation between individuals and
small difference between groups
• Avoid technical variation
o Process sample at same condition as much as possible
o Technician, reagents, time, procedures
• Randomize samples on array
o Avoid confounding technical and biological factors
o Randomly put samples on different array slides and positions

Center for Research Informatics Bioinformatics Core last updated May 2015
 Frequently Asked Questions
´ What if I do not have replicates of data points?
Understand the limitations of un-replicated data! You cannot separate technical variance
from biological variance, thus, the results only apply to the data points sequenced but cannot
be extrapolated to the population.

´ What is difference between Biological replicates and technical replicates?

Technical replicates: measure quantity from 1 source. This measures the reproducibility of the
results. The differences are based only on technical issues in the measurement. (I weigh
myself three times, do I get different weights? How different?)
Biological replicates: measure a quantity from difference sources under the same conditions.
Tumors from 5 different people with lung cancer may show similar gene expression
patterns. These replicates are useful to show what is similar in your replicates and how they
are different from a different set of conditions (ie. treated, normal).

Biological variation is intrinsic to all organisms; it may be influenced by genetic or

environmental factors, as well as by whether the samples are pooled or individual. Technical
variation is introduced during the extraction, labeling and hybridization of samples.
Measurement is associated with reading the fluorescent signals, which may be affected by
factors such as dust on the array.

References
1. P. L. Auer and R. W. Doerge. 2010. Statistical design and analysis of RNA sequencing data.
Genetics 185:405-416.

Center for Research Informatics Bioinformatics Core last updated May 2015