Computation-based discovery of cis-regulatory modules by hidden Markov model

Modelling ChIP-seq Data Using HMMs

Chromatin ImmunoPrecipitation-sequencing (ChIP-seq) experiments have now become routine in biology for the detection of protein binding sites. In this chapter, we show how hidden Markov models can be used for the analysis of data generated by ChIP-seq experiments. We show how a hidden Markov model can naturally account for spatial dependencies in the ChIP-seq data, how it can be used in the presence of data from multiple ChIP-seq experiments under the same biological condition, and how it naturally accounts for the different IP efficiencies of individual ChIP-seq experiments.

The orientation of transcription factor binding site motifs in gene promoter regions: does it matter?

Background

Gene expression is to large degree regulated by the specific binding of protein transcription factors to cis-regulatory transcription factor binding sites in gene promoter regions. Despite the identification of hundreds of binding site sequence motifs, the question as to whether motif orientation matters with regard to the gene expression regulation of the respective downstream genes appears surprisingly underinvestigated.

Results

We pursued a statistical approach by probing 293 reported non-palindromic transcription factor binding site and ten core promoter motifs in Arabidopsis thaliana for evidence of any relevance of motif orientation based on mapping statistics and effects on the co-regulation of gene expression of the respective downstream genes. Although positional intervals closer to the transcription start site (TSS) were found with increased frequencies of motifs exhibiting orientation preference, a corresponding effect with regard to gene expression regulation as evidenced by increased co-expression of genes harboring the favored orientation in their upstream sequence could not be established. Furthermore, we identified an intrinsic orientational asymmetry of sequence regions close to the TSS as the likely source of the identified motif orientation preferences. By contrast, motif presence irrespective of orientation was found associated with pronounced effects on gene expression co-regulation validating the pursued approach. Inspecting motif pairs revealed statistically preferred orientational arrangements, but no consistent effect with regard to arrangement-dependent gene expression regulation was evident.

Conclusions

Our results suggest that for the motifs considered here, either no specific orientation rendering them functional across all their instances exists with orientational requirements instead depending on gene-locus specific additional factors, or that the binding orientation of transcription factors may generally not be relevant, but rather the event of binding itself.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-016-2549-x) contains supplementary material, which is available to authorized users.

Identifying Cis-Regulatory Elements and Modules Using Conditional Random Fields

Accurate identification of cis-regulatory elements and their correlated modules is essential for analysis of transcriptional regulation, which is a challenging problem in computational biology. Unsupervised learning has the advantage of compensating for missing annotated data, and is thus promising to be effective to identify cis-regulatory elements and modules. We introduced a Conditional Random Fields model, referred to as CRFEM, to integrate sequence features and long-range dependency of genomic sequences such as epigenetic features to identify cis-regulatory elements and modules at the same time. The proposed method is able to automatically learn model parameters with no labeled data and explicitly optimize the predictive probability of cis-regulatory elements and modules. In comparison with existing methods, our method is more accurate and can be used for genome-wide studies of gene regulation.

Prediction of clustered RNA-binding protein motif sites in the mammalian genome

Sequence-specific interactions of RNA-binding proteins (RBPs) with their target transcripts are essential for post-transcriptional gene expression regulation in mammals. However, accurate prediction of RBP motif sites has been difficult because many RBPs recognize short and degenerate sequences. Here we describe a hidden Markov model (HMM)-based algorithm mCarts to predict clustered functional RBP-binding sites by effectively integrating the number and spacing of individual motif sites, their accessibility in local RNA secondary structures and cross-species conservation. This algorithm learns and quantifies rules of these features, taking advantage of a large number of in vivo RBP-binding sites obtained from cross-linking and immunoprecipitation data. We applied this algorithm to study two representative RBP families, Nova and Mbnl, which regulate tissue-specific alternative splicing through interacting with clustered YCAY and YGCY elements, respectively, and predicted their binding sites in the mouse transcriptome. Despite the low information content in individual motif elements, our algorithm made specific predictions for successful experimental validation. Analysis of predicted sites also revealed cases of extensive and distal RBP-binding sites important for splicing regulation. This algorithm can be readily applied to other RBPs to infer their RNA-regulatory networks. The software is freely available at http://zhanglab.c2b2.columbia.edu/index.php/MCarts.

Hidden Markov model and its applications in motif findings

Hidden Markov models have wide applications in pattern recognition. In genome sequence analysis, hidden Markov models (HMMs) have been applied to the identification of regions of the genome that contain regulatory information, i.e., binding sites. In higher eukaryotes, the regulatory information is organized into modular units called cis-regulatory modules. Each module contains multiple binding sites for a specific combination of several transcription factors. In this chapter, we gave a brief review of hidden Markov models, standard algorithms from HMM, and their applications to motif findings. We then introduce the application of HMM to a complex system in which an HMM is combined with Bayesian inference to identify transcription factor binding sites and cis-regulatory modules.

Genome-wide analysis of binding sites and direct target genes of the orphan nuclear receptor NR2F1/COUP-TFI

Background

Identification of bona fide direct nuclear receptor gene targets has been challenging but essential for understanding regulation of organismal physiological processes.

Results

We describe a methodology to identify transcription factor binding sites and target genes in vivo by intersecting microarray data, computational binding site queries, and evolutionary conservation. We provide detailed experimental validation of each step and, as a proof of principle, utilize the methodology to identify novel direct targets of the orphan nuclear receptor NR2F1 (COUP-TFI). The first step involved validation of microarray gene expression profiles obtained from wild-type and COUP-TFI^−/− inner ear tissues. Secondly, we developed a bioinformatic tool to search for COUP-TFI DNA binding sites in genomes, using a classification-type Hidden Markov Model trained with 49 published COUP-TF response elements. We next obtained a ranked list of candidate in vivo direct COUP-TFI targets by integrating the microarray and bioinformatics analyses according to the degree of binding site evolutionary conservation and microarray statistical significance. Lastly, as proof-of-concept, 5 specific genes were validated for direct regulation. For example, the fatty acid binding protein 7 (Fabp7) gene is a direct COUP-TFI target in vivo because: i) we identified 2 conserved COUP-TFI binding sites in the Fabp7 promoter; ii) Fapb7 transcript and protein levels are significantly reduced in COUP-TFI^−/− tissues and in MEFs; iii) chromatin immunoprecipitation demonstrates that COUP-TFI is recruited to the Fabp7 promoter in vitro and in vivo and iv) it is associated with active chromatin having increased H3K9 acetylation and enrichment for CBP and SRC-1 binding in the newborn brain.

Conclusion

We have developed and validated a methodology to identify in vivo direct nuclear receptor target genes. This bioinformatics tool can be modified to scan for response elements of transcription factors, cis-regulatory modules, or any flexible DNA pattern.