Bayesian inference based modelling for gene transcriptional dynamics by integrating multiple source of knowledge

Wang Shu-Qiang, Li, Li Han-Xiong

Description

A key challenge in the post genome era is to identify genome-wide transcriptional regulatory networks, which specify the interactions between transcription factors and their target genes. Numerous methods have been developed for reconstructing gene regulatory networks from expression data. However, most of them are based on coarse grained qualitative models, and cannot provide a quantitative view of regulatory systems. Results A binding affinity based regulatory model is proposed to quantify the transcriptional regulatory network. Multiple quantities, including binding affinity and the activity level of transcription factor (TF) are incorporated into a general learning model. The sequence features of the promoter and the possible occupancy of nucleosomes are exploited to estimate the binding probability of regulators. Comparing with the previous models that only employ microarray data, the proposed model can bridge the gap between the relative background frequency of the observed nucleotide and the gene's transcription rate. Conclusions We testify the proposed approach on two real-world microarray datasets. Experimental results show that the proposed model can effectively identify the parameters and the activity level of TF. Moreover, the kinetic parameters introduced in the proposed model can reveal more biological sense than previous models can do.

Informations

• Publié par: biomed
• Publié le: 01 janvier 2012
• Nombre de lectures: 5
• Langue: English
• Poids de l'ouvrage: 1 Mo

Extrait

Wang and Li BMC Systems Biology 2012, 6(Suppl 1):S3
http://www.biomedcentral.com/1752-0509/6/S1/S3
RESEARCH Open Access
Bayesian inference based modelling for gene
transcriptional dynamics by integrating multiple
source of knowledge
1,2 1,2*Shu-Qiang Wang , Han-Xiong Li
From The 5th IEEE International Conference on Computational Systems Biology (ISB 2011)
Zhuhai, China. 02-04 September 2011
Abstract
Background: A key challenge in the post genome era is to identify genome-wide transcriptional regulatory
networks, which specify the interactions between transcription factors and their target genes. Numerous methods
have been developed for reconstructing gene regulatory networks from expression data. However, most of them
are based on coarse grained qualitative models, and cannot provide a quantitative view of regulatory systems.
Results: A binding affinity based regulatory model is proposed to quantify the transcriptional regulatory network.
Multiple quantities, including binding affinity and the activity level of transcription factor (TF) are incorporated into
a general learning model. The sequence features of the promoter and the possible occupancy of nucleosomes are
exploited to estimate the binding probability of regulators. Comparing with the previous models that only employ
microarray data, the proposed model can bridge the gap between the relative background frequency of the
observed nucleotide and the gene’s transcription rate.
Conclusions: We testify the proposed approach on two real-world microarray datasets. Experimental results show
that the proposed model can effectively identify the parameters and the activity level of TF. Moreover, the kinetic
parameters introduced in the proposed model can reveal more biological sense than previous models can do.
Background In recent years, researchers have proposed many dif-
A challenge facing molecular biology is to develop quan- ferent computational approaches to reconstruct gene
titative, predictive models of gene regulation. The regulatory networks from high-throughput data, e.g. see
advance of high-throughput microarray technique reviews by Bansal et al. and Markowetz and Spang [1,2].
makes it possible to measure the expression profiles of These approaches fall roughly into two categories: quali-
thousands of genes, and genome-wide microarray data- tative and quantitative aspects. Inferring qualitative reg-
sets are collected, providing a way to reveal the complex ulatory networks from microarray data has been well
regulatory mechanism among cells. There are two broad studied, and a number of effective approaches have been
classes of gene regulatory interactions: one based on the developed [3-10]. However, these methods are based on
‘physical interaction’ that aim at identifying relationships coarse grained qualitative models [11,12], and cannot
among transcription factors and their target genes provide a realistic and quantitative view of regulatory
(gene-to-sequence interaction) and another based on the systems. On the other hand, quantitative modelling for
‘influence interaction’ that try to relate the expression of gene regulatory network is in its infancy. Research on
a gene to the expression of the other genes in the cell quantitative models for genetic regulation has arisen
(gene-to-gene interaction). only in recent years, and most of them are based on
classical statistical techniques. Liebermeister et al. [13]
1 proposed a linear model for cell cycle-related geneDepartment of Systems Engineering and Engineering Management, City
University of Hong Kong, Hong Kong expression in yeast based on independent component
Full list of author information is available at the end of the article
© 2012 Wang and Li; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.Wang and Li BMC Systems Biology 2012, 6(Suppl 1):S3 Page 2 of 13
http://www.biomedcentral.com/1752-0509/6/S1/S3
analysis. Holter et al. [14] employ singular value decom- reflect time after lights on in hours. For details, please
position to uncover the fundamental patterns underlying refer to Table S1 in additional file 1.
gene expression profiles. Pournara et al. [15] and Yu et
al. [16] proposed the Factor analysis model to describe a Analysis of the predicted activity levels of transcription
larger number of observed variables. However, these factors
approaches are based on linear regression, and are not To test the proposed model on the above dataset, we
always being consistent with the observations in bio- employ two important transcriptional regulators of
chemical experiments which are nonlinear. Imoto et al. which activity levels indicate the variation of heat signals
[17] proposed a nonlinear model with heterogeneous in a subset of gene circadian network, hsf1 and ppara. In
error variances. This model matches the microarray data total, we selected 7 genes to perform posterior inference
wellbutitisnotsatisfying enough in revealing more of TF activities. To ensure identifiability, we included
biological sense. Segal et al. [18] proposed a transcrip- three genes that are regulated solely by hsf1 (HSP110,
tion control network based model and apply their HSPA8 and COL4A1), and two genes that are regulated
model to the segmentation gene network of Drosophila solely by ppara (ACSL1 and HMGCS1). The remaining
melanogaster. They reveal that positional information is two genes are jointly regulated by hsf1 and ppara. These
encoded in the regulatory sequence and input factor dis- genes were chosen since they exhibit the largest variance
tribution. However, there is still a little bit of dilemma in the microarray time course, and hence are likely to
in the model: the activity level of transcription factors is provide a cleaner representation of the output of the
difficult to be measured or to be identified. Actually, system.
assaying the transcription factors’ activity state in a The inferred TFs’ activity levels are shown in Figure 1
dynamic fashion is a major obstacle to the wider appli- (a) and 1(b). Both inferred TF profiles show a noisy per-
cation of the kinetic modeling. TFs’ activity levels are iodic behaviour [20]. Figure 1(c) gives the values of the
difficult to measure mainly due to two technical limita- parameters k for the four selected circadian genesi
tions: TFs are often present at low intercellular concen- (HSPA8, ACSL1, HSP90AA1 and HSPA1B). The green
trations and the changes in their activity state can occur column represents the response k to hsf1 alone, the red1
rapidly due to post-translational modifications. is the response k to ppara alone and the black2
Based on the above description, this paper aims to column represents the joint response k . It can be seen12
describe the transcriptional regulatory network quantita- that, for gene, HSPA8, the model predicts a clear activa-
tively. In this work, a Bayesian inference based regula- tion by hsfl alone, which is consistent with the experi-
tory model is proposed to quantify the transcriptional mental conclusion from Yan et al [20]. The black
dynamics. Multiple quantities, including binding energy, columns of HSP90AA1 and HSPA1B demonstrate that
binding affinity and the activity level of transcription the model predicts a significant combinatorial activation
factor are incorporated into a general learning model. which can be verified by mutagenetic techniques, i.e. by
The sequence features of the promoter and the occu- knocking out one of the TFs.
pancy of nucleosomes are exploited to derive the bind-
ing energy. Compared with the previous models, the The biological sense of kinetic parameters
proposed model can reveal more biological sense. Table 1 shows the relationship between scaling para-
meter k and the corresponding binding affinity .In
Results table 1, ‘H’ indicates ‘high’ and ‘L’ indicates ‘low’.We
Case Ι: Circadian patterns in rat liver define the scaling parameter ki as ‘High’ if it is bigger
Circadian rhythm is a daily time-keeping mechanism than the mean value, as ‘low’,otherwise,andthesame
fundamental to a wide range of species. The basic mole- to binding affinity . From Table 1, we can find that, for
cular mechanism of circadian rhythm has been studied most case, the scaling parameter is in accordance with
extensively. As a real example to test our approach, we the binding affinity: High scaling parameter coupling
considered the dynamics of the circadian patterns in rat with high binding affinity, vice versa. However, gene
liver. We employ the datasets from Almon et al [19]. COL4A1 and HSP110 are 2 exceptions: they have high
This experiment was designed to examine fluctuations scaling parameter but low binding affinity. Our view is
in gene expression in liver within the 24 hour circadian that low binding affinity but high value for k mighti
cycle in normal animals. Fifty-four male normal Wistar represent a TF which rarely binds to promoter but can
rats were housed in a strictly controlled stress free strongly regulate gene expression when bound.
environment with light: dark cycles of 12 hr: 12 hr. Figure 2 shows the results of inference on the values
Three animals were sacrificed at each of 18 selected of the parameters c and ω. The columns on the left,j j
time points within the 24 hour cy