Translating ceRNA susceptibilities into correlation functions

Araks Martirosyan,1, 2 Matteo Marsili,3 and Andrea De Martino4, 5

1
Dipartimento di Fisica, Sapienza Universit`a di Roma, Rome, Italy
VIB-KU Leuven Center for Brain & Disease Research, Leuven, Belgium
3
The Abdus Salam International Centre for Theoretical Physics, Trieste, Italy
4
Soft and Living Matter Lab, Institute of Nanotechnology (CNR-NANOTEC), Consiglio Nazionale delle Ricerche, Rome, Italy
5
Human Genetics Foundation, Turin, Italy

arXiv:1701.05400v1 [q-bio.MN] 19 Jan 2017

Competition to bind microRNAs induces an effective positive crosstalk between their targets, therefore known
as competing endogenous RNAs or ceRNAs. While such an effect is known to play a significant role in specific
conditions, estimating its strength from data and, experimentally, in physiological conditions appears to be far
from simple. Here we show that the susceptibility of ceRNAs to different types of perturbations affecting their
competitors (and hence their tendency to crosstalk) can be encoded in quantities as intuitive and as simple to
measure as correlation functions. We confirm this scenario by extensive numerical simulations and validate
it by re-analyzing PTENs crosstalk pattern from TCGA breast cancer dataset. These results clarify the links
between different quantities used to estimate the intensity of ceRNA crosstalk and provide new keys to analyze
transcriptional datasets and effectively probe ceRNA networks in silico.

INTRODUCTION

MicroRNAs (miRNAs) are small non coding RNA

(ncRNA) molecules that post-transcriptionally regulate a significant portion of the eukaryotic transcriptome via sequencespecific, protein-mediated binding in the cytoplasm [1]. Their
primary effects on coding transcripts consist in inhibiting
translation and fostering degradation [2]. Long ncRNAs,
instead, can transiently sequester miRNAs, thereby altering
their availability and overall repressive potential [3]. Following early observations concerning small regulatory RNAs in
plants and bacteria [4, 5], competition to bind miRNAs has
been hypothesized to cause an effective positive interaction
(crosstalk) between their coding and/or non-coding targets
that may directly affect protein levels [6] (see Fig. 1A,B).
Several experimental and modeling studies have clarified the
conditions under which such a scenario may become biologically relevant, highlighting specifically how molecular levels and kinetic heterogeneities may control it [1, 813]. So
far, such a ceRNA effect (whereby ceRNA stands for competing endogenous RNA) has been quantitatively validated in
cases of differentiation [14], disease [15] or in presence of
unphysiologically large transcriptional inputs [16]. Its significance in standard physiological conditions is therefore subject
to scrutiny [17].
A major difficulty in detecting the ceRNA effect unambiguously in experiments or data lies in the fact that it should be
disentangled from other mechanisms that may bear a similar impact, i.e. an effective positive coupling, on transcripts.
Imagine a network of N ceRNA species interacting with M
miRNA species. ceRNA levels mi (i = 1, . . . , N ) fluctuate
stochastically in time due to random synthesis and degradation events and to interactions with miRNAs, whose levels
are also subject to random fluctuations. Denoting by hi the
time-average in the steady state, an effective ceRNA-ceRNA
dependence can be signaled by a statistical correlation coefficient such as Pearsons [8], i.e.
ij =

[(hm2i i

hmi mj i hmi i hmj i

,

2

2
hmi i )( m2j hmj i )]1/2

(1)

with the idea that, if ij is large enough, a perturbation altering the level of ceRNA j will cause part of the miRNA population to move from one target to the other, effectively broadcasting the perturbation from ceRNA j to ceRNA i through
miRNA-mediated interactions. A more direct description of
this mechanism is attained instead via susceptibilities like [1]
ij =

hmi i
0 ,
bj

(2)

where bj stands for the transcription rate of ceRNA j. ij

quantifies the shift in the mean level of ceRNA i caused by
a (small) variation in bj , and a large ij (assuming no direct
control of ceRNA i by ceRNA j) points to miRNA-mediated
crosstalk between ceRNAs i and j (see Fig. 1C).
While both ij and ij capture aspects of ceRNA crosstalk
seen in experiments, their underlying physical meaning is a
priori different. Fluctuating miRNA levels naturally correlate co-regulated targets, so that a large ij is obtained when
both ceRNAs respond to the stochastic dynamics of their regulator. This however does not necessarily imply a large ij .
In fact, ij can be large even in absence of fluctuations in
miRNA levels, i.e. as a consequence of competition alone. In
such conditions, ij vanishes. ij has indeed been found to
be asymmetric under exchange of its indices (i.e. ij 6= ji
in general) [1], at odds with ij which is necessarily symmetric. It would therefore be important to clarify how quantities
like (1) and (2) are related in miRNA-ceRNA networks, especially to understand whether responses to perturbations (a
central quantity of interest for many potential applications of
the ceRNA effect) can be encoded in quantities as intuitive and
as simple to measure experimentally or from data as a Pearson
correlation coefficient.
Here we show that the information conveyed by ij is indeed captured by a correlation function similar to ij . On the
other hand, ij is linked to a susceptibility, i.e. to the response
of a target to a perturbation altering the level of its competitor, but the perturbation concerns the intrinsic decay rate of
the competitor rather than its transcription (as is the case for
ij ). In the following, we will derive these results and vali-

(A)

(B)

miRNA

target j

target i

mj

induces mi

FIG. 1. (A) A miRNA species interacts with two different target RNAs who are competing to bind it, partially repressing their pools (repression
ways
quantify
the ceRNA
: of one of the targets can induce a de-repression of the
being representedDifferent
by red crosses).
(B) to
In specific
conditions,
an increase effect
in the level
competitor, thereby establishing an effective positive coupling between the targets. Note that a similar effect can in principle be obtained
between ceRNAs that are not co-regulated by the same miRNA species through chains of miRNA-mediated interactions.

(C)

ij>0

ij>0

fluctuations of mi and
mi increases
upon
date them by computer simulations and
gene expression
data
Supporting Text)
mj their
are consequences.
statistically
increasing target js
analysis, and explore
correlated
transcription rate
RESULTS
Theory

We start from the dynamics of molecular populations in a

miRNA-ceRNA network, denoting by mi the level of ceRNA
species i (ranging from 1 to N ), by a the level of miRNA
species a (ranging from 1 to M ), and by cia the levels of
miRNA-ceRNA complexes. In the deterministic limit where
stochastic fluctuations are neglected, the time evolution of
concentration variables is described by
X
X
+

a = a a a
kia
mi a +
(kia
+ ia )cia
i

m
i = bi d i m i

+
kia
mi a

a
+
cia = kia
mi a

kia
cia

(3)

cia
,
ia

with the different parameters denoting intrinsic synthesis

(bi , a ) and degradation rates (di , a ), complex associa
tion/dissociation rates (kia
) and complex processing rates (ia
and ia for stoichiometric and catalytic processing, respec 1
tively), while ia = (ia + ia + kia
) represents the mean
lifetime of the complex formed by miRNA a and ceRNA i.
We note that if the mean lifetime of complexes is much shorter
than that of free molecular species, i.e. if ia  1/di and
ia  1/a for each i and a, miRNA-ceRNA complexes
achieve a steady state much faster than miRNA and ceRNA
levels. In such conditions, cia ' 0 and one can eliminate
complexes from (3) by replacing cia with its steady state value
+
hcia i = kia
ia mi a .

(4)

For ia + kia
 ia (i.e. when stoichiometric degradation
without miRNA recycling is the dominant channel of complex processing), this allows to re-cast (3) in the form (see

BUT

ij>0

ij>0
L
m
i ' mi
,
mi
L
a ' a
,
a

(5)

where L is a function of all miRNA levels = {a } and all

ceRNA levels m = {mi } given by
L=

(bi log mi di mi )

X
X
+
(a log a a a ) +
kia
mi a . (6)
a

i,a

One easily sees (see Supporting Text) that L decreases along

trajectories of (23), implying that its minimum describes the
physically relevant steady state of (3) with m 6= 0 and 6= 0.
If intrinsic molecular noise (arising from stochastic transcription and degradation events and from titration due to
miRNA-ceRNA interactions) is added to (3), after a transient
molecular levels will eventually stabilize and fluctuate over
time around the steady state described by the minimum of L.
We are interested in finding a compact and intuitive mathematical form for the correlations arising between the different
components in such conditions. Molecular noise is Poissonian, namely the strength of fluctuations affecting each variable is proportional to the square root of mean molecular levels (see e.g. [12] for an explicit representation in the context of
a miRNA-ceRNA network), which makes our goal especially
challenging. However we will see that the effects of molecular noise can be remarkably well approximated by a uniform
effective temperature T representing the strength of fluctuations affecting all molecular species involved. In this case,
one can describe fluctuations around the steady state as thermal fluctuations around a Boltzmann-Gibbs equilibrium state.
This allows to compute averages of generic functions of m
and as thermal averages, i.e.
hf i =

1 X
f (m, )eL/T ,
Z(T ) m,

(7)

3
where
Z(T ) =

eL/T

(8)

m,

is a normalization factor, the deterministic limit being obtained for T 0. In particular, defining
hf gic hf gi hf i hgi ,

(9)

by straightforward calculations one finds

log Z(T ) ,
di
hmi i
hmi mj ic = T
T ij ,
dj
hmi log mj ic = T ij .
hmi i = T

(10)

miRNA-ceRNA complex processing dominated by the stoichiometric channel, with the former playing the key role in deriving the function L (see Supporting Text). We note however
that the overall scenario just described also holds for when
complexes evolve over time scales much longer than those of
free molecular levels, i.e. for ia  1/di and ia  1/a .
In particular, (3) can again be re-cast in the form of (23) with
L given by (22), albeit with re-scaled transcription rates (see
Supporting Text for details).
Therefore we conclude that, as long as molecular noise can
be approximated by a uniform effective temperature,
(i) the ceRNA-ceRNA covariance Cij = hmi mj ic is a
proxy for the susceptibility ij , and

(11)
(12)

(ii) the correlation function Xij = hmi log mj ic is a proxy

for the susceptibility ij .

Therefore, in this approximation, the susceptibility ij [Eq.

(2)] is linked to the correlation function [Eq. (12)]
Validation

Xij = hmi log mj ic

(13)

which, as ij , is not symmetric under the exchange of i and j,

while the ceRNA-ceRNA covariance
Cij = hmi mj ic

(14)

is tied to the susceptibility ij quantifying the change in hmi i

induced by a (small) change of the intrinsic degradation rate
dj of ceRNA j [Eq. (11)]. (Note that ij 0.) The constant
linking these quantities is the temperature T quantifying the
strength of the uniform effective noise.
Somewhat unexpectedly, the above results suggest that the
susceptibility ij must be symmetric under exchange of i and
j, i.e., for instance, if the level of ceRNA i is altered by changing the intrinsic degradation rate of ceRNA j, then the reverse
is also true. To check this property, one can calculate ij explicitly for a system formed by N ceRNA species interacting
with a single miRNA species at steady state by following a
different route, specifically along the lines of [1]. Considering
the repression strength to which ceRNAs i and j are subject
at a given (mean) level h1 i of miRNA species 1 (M = 1
in this case), one finds, for each ceRNA, a soft threshold
+
value of h1 i, denoted by 0,i1 ' di /ki1
, such that i is unrepressed (resp. repressed or susceptible to changes in h1 i) if
h1 i  0,i1 (resp.  0,i1 or ' 0,i1 ). A direct calculation
(see Supporting Text) shows that ij can attain large values
only if both ceRNAs are susceptible to h1 i, in which case
one has (i 6= j)
!1
X b` k +
h1 i bi bj
`1
ij '
1 +
a`
, (15)
0,i1 0,j1 di dj
d`
`

where a` = 0, 1, 1/4 if ceRNA ` is repressed, unrepressed or

susceptible, respectively. Eq. (15) confirms that ij is indeed
symmetric under exchange of i and j.
Concerning the approximations under which the the above
results were obtained, we remark that we started by considering (3) in the limit of (i) fast complex equilibration, and (ii)

We have validated the above scenario by simulating a small

network involving 2 ceRNA and a single miRNA species
via the Gillespie algorithm [2], where molecular noise is accounted for explicitly (see Supporting Text). Results are summarized in Fig. 2, where we compare 12 , 21 and C12 C21
on on hand, and 12 , 21 , X12 and X21 on the other, as computed from simulations (i.e. with the actual molecular noise),
against the theoretical predictions. We considered three scenarios for the mean lifetime of miRNA-ceRNA complexes,
namely those covered by the theory (i.e. complex equilibration much faster and much slower than the equilibration of
miRNA and ceRNA levels) as well as the intermediate case
where characteristic timescales are comparable for all variables.
One sees that theoretical predictions obtained in the thermal noise approximation agree remarkably well with simulations including the actual molecular noise. In particular, the
full correspondence between the susceptibilities ij and ij
and the (re-scaled) correlation functions Cij and, respectively,
Xij is evident. Notice that a single global parameter T 0
has been used to fit all data in each of the conditions. This
shows how accurately the assumption of a uniform effective
temperature can mimic the effects of intrinsic stochasticity.
On the other hand, its limits might be reflected, at least in
part, in the discrepancies that occur at high transcription rates.
Parameters [units]

Figs 2A Figs 2B Figs 2C

b2 , 1 [molec. min1 ]
10
+
k11
[molec.1 min1 ] e5
+
k21
[molec.1 min1 ] e6

k11
, k21
[min1 ]
0.001
d1 , d2 , 1 [min1 ]
0.1
11 , 21 [min1 ]
1
11 , 21 [min1 ]
0.001

10
e5
e6
0.001
0.05
0.05
0.001

10
e5
e6
0.001
0.2
0.05
0.001

TABLE I. Parameters used in Fig 2. Note that i {1, 2} while

a = 1.

<< d-1

10

20
b1

30

40

1.0

12

0.5

X12/T

21

0.0

12

1.5

-1

<< d

1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4

10

20
b1

30

40

6
5
4
3
2
1
0

21

<< d-1

X21/T

10

20
b1

21

12

2.0

30

100

-21
-12
C21/T

-21
-12
C21/T

100
6
5
4
3
2
1
0
0

= d -1

10

20
b1

30

40

12
X12/T
0

10

0.25
0.20
0.15
0.10
0.05
0.00

12

1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0

(C)

1/,1/d
T 0.09

20
b1

= d -1

30

40

21

7
6
5
4
3
2
1

= d -1

X21/T

10

20
b1

30

21

100

(B)

1/,1/d
T 0.14

(A)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0

1/,1/d
T 0.2

-21
-12
C21/T
>> d-1

10

20
b1

30

40

12
X12/T

>> d-1

10

20
b1

30

21

>> d-1

0.35
0.30
0.25
0.20
0.15

X21/T
0

10

20
b1

30

FIG. 2. Results from Gillespie simulations (markers) versus analytical predictions for susceptibilities (lines) for a miRNA-ceRNA network with
N = 2 ceRNA and M = 1 miRNA species and different values of the transcription rate of ceRNA species 1, b1 . Numerical susceptibilities
have been obtained by perturbing the system at steady state and recording the new steady state at which the system settles after a transient.
Analytical lines correspond to the numerical derivatives of the steady state levels obtained from (3). Sets (A)(C) are for different values of the
mean lifetime of miRNA-ceRNA complexes , and refer, respectively, to the cases in which complex processing is much faster, comparable
or much slower, than the degradation dynamics of free miRNAs and ceRNAs. Also reported is the value of T that provides the best fit in each
case. See Table I for parameter values. Results obtained for stronger miRNA-ceRNA couplings are shown in the Supporting Text.

These results confirm that (12) and (14) are indeed good
predictors of the response of a ceRNA to a perturbation affecting one of its competitors within a miRNA-ceRNA network. Notably, such correlation functions are easy to estimate
from transcription data sets. Our framework therefore has the
potential to offer new insight into post-transcriptional regulation, its system-level organization and its impact on cellular
functions.
In order to test this idea, we analyzed the ceRNA scenario
emerging from 1098 breast cancer samples obtained from The
Cancer Genome Atlas [19], focusing on the widely studied
oncosuppressor PTEN and its immediate competitors (i.e. the
ceRNAs sharing at least one miRNA regulator with PTEN).
In particular, we computed
CPTEN,ceRNA = hmPTEN mceRNA ic
XceRNA,PTEN = hmceRNA log mPTEN ic
XPTEN,ceRNA = hmPTEN log mceRNA ic

(16)
(17)
(18)

for a set of candidate PTEN ceRNAs found in [15] by means

of Mutually Targeted miRNA-Response Element Enrichment

Analysis. Notice that the average appearing in Eqs. (16

18) is over samples and not over time. We expect however
that, if the interaction network is conserved across samples,
averages over samples should reproduce statistical averages
such as (13), as different samples effectively represent different snapshots of the state of the network. Fig. 3A shows that
when (16) (whose value is encoded in the color of markers) is
large, both (17) and (18) tend to be large. According to (15),
a large Cij (or ij ) signals that both PTEN and its competitor
are susceptible to changes in the level of at least one of their
shared regulators. For such pairs, in addition, it has previously
been shown that both ij and ji are expected to be large [1].
This implies a fully bi-directional crosstalk, i.e. any perturbation affecting the level of one species should affect the level of
the other via miRNA-mediated regulation. Remarkably, this
was experimentally shown to be the case in [15] for some of
the ceRNAs we tested (e.g. SERINC1, VAPA), all of which
are in this regime according to our analysis. Adding to this,
we are also able to point to a number of other PTEN competitors, a perturbation of which should trigger a response by

TE
N

ESR1

RN

TE
N

,ce

RN

RAPH1

ZEB2

1.5

DTWD2
PRELP PIK3C2A
MS4A7

C11ORF63

1.3

SERINC1

SMAD5
QKI

ZNF175

SEPT11

ANKRD36C
ANKRD36B

1.1

GALNT16

NUBPL
ZFP14
KIAA1958
IL6R
CD28

0.9

KIAA0391
AMY2B
NUDT13

0.7
0.5

NBR1

KCTD20
SLC1A2

GNB5
FAM217B
NRIP2

0.0

VCAN

VKORC1L1

PRKCA

PAK2
WDR82

IFNAR1

1.0

RAPH1

ZEB2

1.5

DTWD2

C11ORF63

1.3

ZNF175

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

SERINC1

SMAD5
QKI
SEPT11

ANKRD36B

GALNT16
NUBPL
ZFP14
VKORC1L1
PRKCA
KCTD20
KIAA1958
SLC1A2 IFNAR1
PAK2
IL6R
WDR82
CD28
GNB5
KIAA0391
FAM217B
FAM84B
AMY2B
NRIP2

1.1

NUDT13

0.7

COV(ceRNA, log PTEN) x 10

XceRNA,PTEN
(1000)

PRELP PIK3C2A
MS4A7

ANKRD36C

0.9

HACD3
MLEC

VAPA

0.5

XPTEN,ceRNA XceRNA,PTEN

1.7

RNF213

FAM84B

0.6

1.9

A=

1.7

0.4

ESR1

Xce

XPTEN,ceRNA XceRNA,PTEN

XP

XPTEN,ceRNA
(1000)
COV(PTEN,
log ceRNA)
x 103

1.9

PTEN,ceRNA
Pearson(PTEN,
ceRNA)
0.2

2.1

A,P

2.1

(B)

7
CCOV(PTEN,
(10
PTEN,ceRNA
ceRNA)
x 10)7
1

X
(1000)
PTEN,ceRNA
COV(PTEN,
log ceRNA)
x 103

(A)

0.5

0.0

VAPA

0.5

1.0

VCAN
NBR1

RNF213
HACD3
MLEC

1.5
2.0
2.5
3.0
COV(ceRNA, log PTEN) x 103
X
ceRNA,PTEN (1000)

3.5

4.0

4.5

5.0

FIG. 3. Values of different correlation functions of PTEN with its ceRNAs computed from breast cancer samples from TCGA. (A) XceRNA,PTEN
and XPTEN,ceRNA are reported on the x and y axes respectively, while the color code gives the value of CPTEN,ceRNA . (B) Same as (A), except
that the color code now corresponds to the Pearson coefficient PTEN,ceRNA = Cij /(mi mj ). Notice that in (B) PTEN-ceRNA pairs occupy
the same positions in the plane as in (A). See Supporting Text for higher-resolution versions of both panels.

PTEN.
On the other hand, smaller values of (16) (orange markers in Fig. 3A) are associated to strongly asymmetric PTENceRNA pairs for which (18) is much larger than (17). This
suggests that PTEN will respond to an increase of its competitors bare transcription rate (and not vice-versa), while no response of PTEN should be expected upon perturbing the bare
decay rate of the same ceRNA as Cij is small. Within the
steady state theory of [1], ceRNA pairs with strongly different values of ij and ji pertain to cases where the responding
ceRNA (PTEN here) is susceptible to variations in the miRNA
levels while the perturbed one (PTENs competitor) is fully
repressed. Our data analysis fully confirms both this scenario
and the theory presented here in linking such cases to low values of the bare covariance (14).
Finally, note (Fig. 3B) that the above information can not
be retrieved if Cij is replaced by the Pearson coefficient ij ,
Eq. (1), which just amounts to normalizing Cij by the product of the standard deviations mi and mj of mi and mj .
Indeed, using the value of ij to color-code PTENs ceRNAs,
one sees that the Pearson coefficient can mislead into expecting (or not expecting) a response to a perturbation when the
actual susceptibilities are small (resp. large).
For instance, ij is rather small for the pair formed by
PTEN and SLC1A2, which seems to suggest absence of mutual cross-talk between these two transcripts. However, while
both CPTEN,SLC1A2 and XSLC1A2,PTEN are small, XPTEN,SLC1A2
is significant. This suggests that (i) SLC1A2 will not respond
to a perturbation affecting the transcription rate of PTEN, and
(ii) the pair will be insensitive to changes in each others
bare decay rate; however, (iii) PTEN will be affected by a
change in the bare transcription rate of its competitor despite
the small statistical correlation that exists between their levels. Likewise, the large value of the Pearson coefficient between PTEN and DTWD2 can mislead into generically expecting a response when instead the susceptibility is strongly
perturbation-dependent. In particular, the level of DTWD2

should not be significantly modified by a change in the level of

PTEN (as XDTWD2,PTEN is rather small) in spite of the large
Pearson coefficient. Notice that, remarkably, for this pair, Cij
and ij take on very different values.
DISCUSSION

To sum up, we have identified [Eq.s (11) and (12)] a set

of correlation functions that can serve as proxies for ceRNA
susceptibilities to perturbations. Specifically, Cij = hmi mj ic
is related to the susceptibility ij quantifying ceRNA is response to a change of the bare decay rate of ceRNA j, while
Xij = hmi log mj ic is related to the susceptibility ij quantifying ceRNA is response to a change of the bare transcription
rate of ceRNA j. These relations are valid at steady state and
within the approximations discussed, are fully confirmed by
numerical simulations.
Most importantly, quantities like Cij and Xij can be easily estimated from data and possibly measured in experiments. An analysis of PTENs emergent crosstalk pattern
from TCGA breast cancer dataset using these functions has
indeed shown that a map of ceRNA responses to perturbations
affecting competitors can be constructed by combining the information provided by each, while the Pearson coefficient ij
can be inaccurate in this respect. This opens the way to probing the structure and function of ceRNA networks in silico by
straightforwardly analyzing transcriptional data, and provides
a key to obtain testable transcriptome-scale predictions about
ceRNA crosstalk.
Notice that our results apply without any modification to
ceRNA pairs that dont share miRNA regulators, i.e. it is capable of identifying long-range crosstalk (i.e. interactions between ceRNAs that are separated by multiple miRNAs along
the miRNA-ceRNA network) of the kind discussed in [20].
From the viewpoint of physics, results like (11) and (12)
are akin to the fluctuation-response relations that constitute

6
a cornerstone of statistical mechanics [21]. Their derivation
in our context has relied on an equilibrium framework that
presupposes stationarity of molecular levels. Since ceRNA
crosstalk can be substantially more complex away from the
steady state [22], a more refined mathematical study will
be required to extend the theory developed here to offequilibrium dynamical regimes. Our results on the other hand
may also open the way to the application of recently devel-

oped inference techniques [23] to estimate miRNA levels or

kinetic parameters from ceRNA levels.

[1] Flynt AS, Lai EC. Biological principles of microRNA-mediated

regulation: shared themes amid diversity. Nat Rev Genet (2008)
9:831-42
[2] Jonas S, Izaurralde E. Towards a molecular understanding of
microRNA-mediated gene silencing. Nat Rev Genet (2015)
16:421-33
[3] Guil S, Esteller M. RNA-RNA interactions in gene regulation:
the coding and noncoding players. TIBS (2015) 40:248-56
[4] Franco-Zorrilla JM, Valli A, Todesco M, Mateos I, Puga MI,
Rubio-Somoza I, Leyva A, Weigel D, Garcia JA, Paz-Ares J.
Target mimicry provides a new mechanism for regulation of microRNA activity. Nat Genet (2007) 39:1033-1037
[5] Levine E, Zhang Z, Kuhlman T, Hwa T Quantitative Characteristics of Gene Regulation by Small RNA. PLoS Biol (2007) 5(9):
e229
[6] Salmena L, Poliseno L, Tay Y, Kats L, Pandolfi PP. A ceRNA
hypothesis: the Rosetta Stone of a hidden RNA language? Cell
(2011) 146:353-8
[7] Figliuzzi M, Marinari E, De Martino A. MicroRNAs as a selective channel of communication between competing RNAs: a
steady-state theory. Biophys J (2013) 104:1203-13
[8] Bosia C, Pagnani A, Zecchina R. Modelling competing endogenous RNA networks. PLoS One (2013) 8:e66609
[9] Jens M, Rajewsky N. Competition between target sites of regulators shapes post-transcriptional gene regulation. Nat Rev Genet
(2015) 16:113-26
[10] Bosson AD, Zamudio JR, Sharp PA. Endogenous miRNA
and target concentrations determine susceptibility to potential
ceRNA competition. Mol Cell (2014) 56:347-59
[11] Lai X, Wolkenhauer O, Vera J. Understanding microRNAmediated gene regulatory networks through mathematical modelling. Nucleic Acids Res (2016) gkw550
[12] Martirosyan A, Figliuzzi M, Marinari E, De Martino A. Probing the limits to microRNA-mediated control of gene expression.
PLoS Comput Biol (2016) 12:e1004715
[13] Martirosyan A, De Martino A, Pagnani A, Marinari E. ceRNA
crosstalk stabilizes protein expression and affects the correlation
pattern of interacting proteins. Submitted (2016)
[14] Legnini I, Morlando M, Mangiavacchi A, Fatica A, Bozzoni I.
A feedforward regulatory loop between HuR and the long noncoding RNA linc-MD1 controls early phases of myogenesis. Mol
Cell (2014) 53:506-14
[15] Tay Y, Kats L, Salmena L, Weiss D, Tan SM, Ala U, Karreth
F, Poliseno L, Provero P, Di Cunto F, Lieberman J, Rigoutsos
I, Pandolfi PP. Coding-independent regulation of the tumor suppressor PTEN by competing endogenous mRNAs. Cell (2011)
147:344-57
[16] Denzler R, Agarwal V, Stefano J, Bartel DP, Stoffel M. Assessing the ceRNA hypothesis with quantitative measurements
of miRNA and target abundance. Mol Cell (2014) 54:766-76

[17] Denzler R, McGeary SE, Title AC, Agarwal V, Bartel DP, Stoffel M. Impact of microRNA levels, target-site complementarity, and cooperativity on competing endogenous RNA-regulated
gene expression. Mol Cell (2016) 64:565-579
[18] Gillespie DT. Stochastic simulation of chemical kinetics. Annu
Rev Phys Chem (2007) 58:35-55
[19] The TCGA Research Network: http://cancergenome.nih.gov/
[20] Nitzan M, Steiman-Shimony A, Altuvia Y, Biham O, Margalit
H. Interactions between distant ceRNAs in regulatory networks.
Biophys J (2014) 106:2254-66
[21] Marconi UM, Puglisi A, Rondoni L, Vulpiani A. Fluctuationdissipation: response theory in statistical physics. Phys Rep
(2008) 461:111-95
[22] Figliuzzi M, De Martino A, Marinari E. RNA-based regulation:
dynamics and response to perturbations of competing RNAs.
Biophys J (2014) 107:1011-22
[23] Roudi Y, Aurell E, Hertz J. Statistical physics of pairwise probability models. Front Comput Neurosci (2009) 3:22

Acknowledgments

We gratefully acknowledge Carla Bosia and Andrea Pagnani for useful insight and suggestions.

7
SUPPORTING TEXT
I.

DERIVATION OF THE FUNCTION L FOR FAST COMPLEX PROCESSING

1
Starting from Eq. (3) of the Main Text, namely (with ia = (ia + ia + kia
) )
X
X
+

a = a a a
kia
mi a +
(kia
+ ia )cia ,
i

m
i = bi di mi

+
kia
mi a

kia
cia ,

(19)

+
cia = kia
mi a

cia
,
ia

we assume that complexes equilibrate much faster than miRNA and ceRNA levels and substitute cia with its steady state value
+
hcia i = kia
ia mi a . One finds
X
ia
+
a ' a a a
kia mi a ,

+
k
ia
ia
ia
i
(20)
X
ia + ia
+
m
i ' bi d i m i
k
m

.
ia i a

ia + ia + kia
a

For ia + kia
 ia , this reduces to

a ' a a a

m
i ' bi di mi

+
kia
mi a ,

(21)

+
kia
mi a ,

which is easily seen to be equivalent to Eq. (5) in the Main Text, with
X
X
X
+
L=
(bi log mi di mi )
(a log a a a ) +
kia
mi a .
a

II.

(22)

i,a

L DECREASES ALONG THE DYNAMICS

By direct differentiation and using the fact that (see Eq. (5) in the Main Text)
L
m
i
'
,
mi
mi
L
a
'
,
a
a

(23)

one finds
L =

X L
X m i 2 X a 2
X L
m i +
a =

0 .
mi
a
mi
a
a
a
i
i

(24)

In other words, under the approximations discussed above, L decreases along the dynamics of the miRNA-ceRNA network.
Therefore the minimum of L (which is unique by virtue of the concavity of L) describes a steady state of the dynamics (19).
III.

APPROXIMATE CALCULATION OF ij FOR A SYSTEM WITH ONE MIRNA AND N CERNA SPECIES

Starting from Eq. (21) taken for N ceRNA species and a single miRNA species (we suppress its index for simplicity) in the
limit i  ki + i , the steady-state level of mi reads
mi =

bi
Fi ()
di

Fi () =

0,i
0,i +

0,i =

di
.
ki+

(25)

8
Now noting that

1 0,i

Fi () ' 21 40,i0,i

0,i

for  0,i (ceRNA i expressed)

0,i
dFi
Fi0 ' 410,i

d
0,i
2

for  0,i (ceRNA i expressed)

for ' 0,i (ceRNA i susceptible)

(26)

for  0,i (ceRNA i repressed)

for ' 0,i (ceRNA i susceptible)

(27)

for  0,i (ceRNA i repressed)

and that the steady state miRNA level can be approximated by [1]
P
P
iRepr bi 14 iSusc bi
,
'
P
P
b k+
b k+
+ iExpr idii + 41 iSusc idii

(28)

so that

for  0,i (ceRNA i expressed)

for ' 0,i (ceRNA i susceptible)
for  0,i (ceRNA i repressed)
1

X bi k +
X bi k +
1
i
i
+
,
= +
d
4
d
i
i
iExpr

bj
'
41
dj
dj 0,j
0

(29)

(30)

iSusc

we can compute the susceptibility ij as

mi
bi

= Fi0
dj
di dj

ij

(i 6= j) .

(31)

One finds
ij '

bi bj
WR(i),R(j)
di dj

(i 6= j) ,

(32)

c is a 33 matrix that only depends on the regime R(i) (repressed, susceptible or expressed) to which ceRNA i belongs.
where W
c are found to be  1 (for
By considering the definitions of the different regimes in terms of the value of , all elements of W
instance, WExpr,Expr = /(0,i 0,j )  1 as  0,i and  0,j if ceRNAs i and j are both expressed) except for
WSusc,Susc , which is given by
WSusc,Susc =

,
16 0,i 0,j

(33)

leading immediately to Eq. (15) of the Main Text.

IV.

CASE OF SLOW COMPLEX PROCESSING

Assuming complex levels cia are roughly stationary over time scales for which mi and a evolve (i.e. ia  1/di and
ia  1/a for each i and a), then all terms in (19) that involve the variables cia can be taken to be roughly constant for short
enough characteristic times. In such conditions, miRNAs are effectively transcribed at rates
X

aeff ' a +
(kia
+ ia )cia ,
(34)
i

while ceRNAs are effectively transcribed at rates

beff
i ' bi +

X
a

kia
cia .

(35)

9
In this limit, (19) can again be cast as
L
,
mi
L
a ' a
,
a

m
i ' mi

(36)

with
L=

(beff
i log mi di mi )

X
X
+
(aeff log a a a ) +
kia
mi a .
a

(37)

i,a

The main difference from the previous case lies in the fact that the minimum of L should now be computed self-consistently
from the asymptotic value of cia : after the (fast) equilibration of mi s and a s following (36), a new steady state value for
+
complexes is computed as cia = kia
ia mi a , leading in turn to new values for the effective transcription rates aeff and beff
i and
hence to new values for mi s and a s from (36), and so on until convergence.
V.

STOCHASTIC DYNAMICS OF A MIRNA-CERNA NETWORK

The time evolution of our miRNA-ceRNA network with N ceRNA species (labeled i), M miRNA species (labeled a) and
intrinsic (molecular) noise is described by the system
X
X
+

a = a a a
kia
mi a +
(kia
+ ia )cia + a ,
i

m
i = bi di mi

+
kia
mi a

kia
cia + i ,

(38)

cia
+ ia ,
ia

+
cia = kia
mi a

1
where ia = (ia + ia + kia
) while a , a and ia represent stochastic variables. As each noise source contributes independently to the overall noise level, one has
X
X
X
+

a = a
ia
+
ia
+
ia
,
(39)
i

i = mi

X
a

+
ia

ia

(40)

ia = ia
+ ia
ia
ia

(41)

where mi , a , ia
, ia
, and ia
and are mutually independent zero-average random variables representing, respectively, the intrinsic noise in ceRNA levels, in miRNA levels, in the binding/unbinding dynamics of complexes, in the stoichiometric complex
degradation channel and in the catalytic complex degradation channel. Correlations are, for each component, described by

hmi (t)mi (t0 )i = (di mi + bi ) (t t0 ) ,

ha (t)a (t0 )i = (a a + a ) (t t0 ) ,

+

+ 0
+
ia (t)ia
(t ) = kia
mi a (t t0 ) ,

ia (t)ia
(t ) = kia
cia (t t0 ) ,

(42)

0
hia
(t)ia
(t )i = ia cia (t t0 ) ,

0
hia
(t)ia
(t )i = ia cia (t t0 ) ,

where
P
bi + a kia
cia
,
P +
di + a kia
a
P
a + i (kia
+ ia )cia
a =
,
P +
a + i kia mi
mi =

cia =

+
kia
a mi

ia + kia
+ ia

(43)
(44)
(45)

10
denote the mean steady-state molecular levels. To obtain Fig. 2 of the Main Text, we have simulated the above system with
M = 1, N = 2 using the Gillespie algorithm [2].

[1] Figliuzzi M, Marinari E, De Martino A. MicroRNAs as a selective channel of communication between competing RNAs: a steady-state
theory. Biophys J (2013) 104:1203-13
[2] Gillespie DT. Stochastic simulation of chemical kinetics. Annu Rev Phys Chem (2007) 58:35-55

30

= -

(C)
1/,1/d
T 0.27

100
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0

C21/T
10

20
b1

21

0.0

30

5
4
3
2
1
0

X12/T
0

10

20
b1

30

10

20
b1

X21/T

10

20
b1

30

= -

20
15
10

X21/T

X12/T
0

<< d -1

= -

0.8

30

>> -

10

20
b1

30

>> -

1.5

0.6

12

1/,1/d
T 0.08

1.0

21

20
b1

0.5

(B)

10

<< d -1

1.5

C21/T
10

20
b1

>> -

30

21

100
6
5
4
3
2
1
0
0

C21/T

12

1/,1/d
T 0.16

2.0

<< d -1

(A)

12

100
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0

0.4
0.2
0.0

10

20
b1

X21/T

0.5

X12/T
0

1.0

30

10

20
b1

30

FIG. 4. Same as Figure 2 in the Main Text but for stronger miRNA repression. Parameter values are as in Table I of the Main Text except for
+
+
k11
= e3 and k21
= e4 . Note that, for small values of b1 , m1 gets too small to accurately estimate Xij s.

11

XPTEN,ceRNA
(1000)
COV(PTEN,
log ceRNA)
x 103

SMAD5
QKI

MS4A7

PRELP PIK3C2A

ZEB2

7
CCOV(PTEN,
(10
PTEN,ceRNA
ceRNA)
x 10)7

SEPT11
NBR1

RNF213

2.5

3.0
3

3.5

SERINC1

4.0

XPTEN,ceRNA XceRNA,PTEN
HACD3
MLEC

2.0

4.5

ESR1

VCAN

5.0

FIG. 5. Large-scale version of Fig. 3A in the Main Text.

XPTEN,ceRNA XceRNA,PTEN
RAPH1

DTWD2

C11ORF63
ZNF175
ANKRD36C
GALNT16
VKORC1L1

KCTD20
PAK2
IFNAR1
WDR82

VAPA

1.5

COV(ceRNA, log PTEN) x 10

XceRNA,PTEN
(1000)

FAM84B

1.0

NA
,PT
EN
ceR

2.1

1.9

ANKRD36B

PRKCA

NUBPL
ZFP14

SLC1A2

0.5

GNB5
FAM217B
NRIP2

KIAA1958
IL6R
CD28
KIAA0391
AMY2B
NUDT13

0.0

=X
NA
,ce
R
TE
N

1.7

1.5

1.3

1.1

0.9

0.7
0.5

XP

12

X
(1000)
PTEN,ceRNA
COV(PTEN,
log ceRNA)
x 103

2.1

1.9

1.7

1.5

1.3

1.1

0.9

0.7

RAPH1

DTWD2

C11ORF63
ZNF175

0.5

CD28
GNB5
KIAA0391
FAM217B
AMY2B
NRIP2
NUDT13

0.0

ZEB2

0.2

PRELP PIK3C2A
MS4A7
SMAD5
QKI

VAPA

FAM84B

1.0

1.5

0.4

RNF213

2.5

3.0

PTEN,ceRNA
Pearson(PTEN,
ceRNA)

2.0

HACD3
MLEC

SEPT11
ANKRD36C
ANKRD36B
GALNT16
NBR1
NUBPL
ZFP14
VKORC1L1
PRKCA
KCTD20
KIAA1958
SLC1A2 IFNAR1
PAK2
IL6R
WDR82

0.5

COV(ceRNA, log PTEN) x 10

X
ceRNA,PTEN (1000)

3.5

SERINC1

4.0

0.6

4.5

ESR1

VCAN

5.0

FIG. 6. Large-scale version of Fig. 3B in the Main Text.