A biophysical model for analysis of transcription factor interaction and binding site arrangement from genome-wide binding data

Base-resolution models of transcription-factor binding reveal soft motif syntax

The arrangement (syntax) of transcription factor (TF) binding motifs is an important part of the cis-regulatory code, yet remains elusive. We introduce a deep learning model, BPNet, that uses DNA sequence to predict base-resolution chromatin immunoprecipitation (ChIP)-nexus binding profiles of pluripotency TFs. We develop interpretation tools to learn predictive motif representations and identify soft syntax rules for cooperative TF binding interactions. Strikingly, Nanog preferentially binds with helical periodicity, and TFs often cooperate in a directional manner, which we validate using clustered regularly interspaced short palindromic repeat (CRISPR)-induced point mutations. Our model represents a powerful general approach to uncover the motifs and syntax of cis-regulatory sequences in genomics data.

Devil in the details: Mechanistic variations impact information transfer across models of transcriptional cascades

The transcriptional network determines a cell’s internal state by regulating protein expression in response to changes in the local environment. Due to the interconnected nature of this network, information encoded in the abundance of various proteins will often propagate across chains of noisy intermediate signaling events. The data-processing inequality (DPI) leads us to expect that this intracellular game of “telephone” should degrade this type of signal, with longer chains losing successively more information to noise. However, a previous modeling effort predicted that because the steps of these signaling cascades do not truly represent independent stages of data processing, the limits of the DPI could seemingly be surpassed, and the amount of transmitted information could actually increase with chain length. What that work did not examine was whether this regime of growing information transmission was attainable by a signaling system constrained by the mechanistic details of more complex protein-binding kinetics. Here we address this knowledge gap through the lens of information theory by examining a model that explicitly accounts for the binding of each transcription factor to DNA. We analyze this model by comparing stochastic simulations of the fully nonlinear kinetics to simulations constrained by the linear response approximations that displayed a regime of growing information. Our simulations show that even when molecular binding is considered, there remains a regime wherein the transmitted information can grow with cascade length, but ends after a critical number of links determined by the kinetic parameter values. This inflection point marks where correlations decay in response to an oversaturation of binding sites, screening informative transcription factor fluctuations from further propagation down the chain where they eventually become indistinguishable from the surrounding levels of noise.

Enhancer occlusion transcripts regulate the activity of human enhancer domains via transcriptional interference: a computational perspective

Analysis of ENCODE long RNA-Seq and ChIP-seq (Chromatin Immunoprecipitation Sequencing) datasets for HepG2 and HeLa cell lines uncovered 1647 and 1958 transcripts that interfere with transcription factor binding to human enhancer domains. TFBSs (Transcription Factor Binding Sites) intersected by these 'Enhancer Occlusion Transcripts' (EOTrs) displayed significantly lower relative transcription factor (TF) binding affinities compared to TFBSs for the same TF devoid of EOTrs. Expression of most EOTrs was regulated in a cell line specific manner; analysis for the same TFBSs across cell lines, i.e. in the absence or presence of EOTrs, yielded consistently higher relative TF/DNA-binding affinities for TFBSs devoid of EOTrs. Lower activities of EOTr-associated enhancer domains coincided with reduced occupancy levels for histone tail modifications H3K27ac and H3K9ac. Similarly, the analysis of EOTrs with allele-specific expression identified lower activities for alleles associated with EOTrs. ChIA-PET (Chromatin Interaction Analysis by Paired-End Tag Sequencing) and 5C (Carbon Copy Chromosome Conformation Capture) uncovered that enhancer domains associated with EOTrs preferentially interacted with poised gene promoters. Analysis of EOTr regions with GRO-seq (Global run-on) data established the correlation of RNA polymerase pausing and occlusion of TF-binding. Our results implied that EOTr expression regulates human enhancer domains via transcriptional interference.

The Role of Chromatin Accessibility in cis-Regulatory Evolution

Transcription factor (TF) binding is determined by sequence as well as chromatin accessibility. Although the role of accessibility in shaping TF-binding landscapes is well recorded, its role in evolutionary divergence of TF binding, which in turn can alter cis-regulatory activities, is not well understood. In this work, we studied the evolution of genome-wide binding landscapes of five major TFs in the core network of mesoderm specification, between Drosophila melanogaster and Drosophila virilis, and examined its relationship to accessibility and sequence-level changes. We generated chromatin accessibility data from three important stages of embryogenesis in both Drosophila melanogaster and Drosophila virilis and recorded conservation and divergence patterns. We then used multivariable models to correlate accessibility and sequence changes to TF-binding divergence. We found that accessibility changes can in some cases, for example, for the master regulator Twist and for earlier developmental stages, more accurately predict binding change than is possible using TF-binding motif changes between orthologous enhancers. Accessibility changes also explain a significant portion of the codivergence of TF pairs. We noted that accessibility and motif changes offer complementary views of the evolution of TF binding and developed a combined model that captures the evolutionary data much more accurately than either view alone. Finally, we trained machine learning models to predict enhancer activity from TF binding and used these functional models to argue that motif and accessibility-based predictors of TF-binding change can substitute for experimentally measured binding change, for the purpose of predicting evolutionary changes in enhancer activity.

Mechanistic interpretation of non-coding variants for discovering transcriptional regulators of drug response

BACKGROUND

Identification of functional non-coding variants and their mechanistic interpretation is a major challenge of modern genomics, especially for precision medicine. Transcription factor (TF) binding profiles and epigenomic landscapes in reference samples allow functional annotation of the genome, but do not provide ready answers regarding the effects of non-coding variants on phenotypes. A promising computational approach is to build models that predict TF-DNA binding from sequence, and use such models to score a variant's impact on TF binding strength. Here, we asked if this mechanistic approach to variant interpretation can be combined with information on genotype-phenotype associations to discover transcription factors regulating phenotypic variation among individuals.

RESULTS

We developed a statistical approach that integrates phenotype, genotype, gene expression, TF ChIP-seq, and Hi-C chromatin interaction data to answer this question. Using drug sensitivity of lymphoblastoid cell lines as the phenotype of interest, we tested if non-coding variants statistically linked to the phenotype are enriched for strong predicted impact on DNA binding strength of a TF and thus identified TFs regulating individual differences in the phenotype. Our approach relies on a new method for predicting variant impact on TF-DNA binding that uses a combination of biophysical modeling and machine learning. We report statistical and literature-based support for many of the TFs discovered here as regulators of drug response variation. We show that the use of mechanistically driven variant impact predictors can identify TF-drug associations that would otherwise be missed. We examined in depth one reported association-that of the transcription factor ELF1 with the drug doxorubicin-and identified several genes that may mediate this regulatory relationship.

CONCLUSION

Our work represents initial steps in utilizing predictions of variant impact on TF binding sites for discovery of regulatory mechanisms underlying phenotypic variation. Future advances on this topic will be greatly beneficial to the reconstruction of phenotype-associated gene regulatory networks.

ChIPulate: A comprehensive ChIP-seq simulation pipeline

ChIP-seq (Chromatin Immunoprecipitation followed by sequencing) is a high-throughput technique to identify genomic regions that are bound in vivo by a particular protein, e.g., a transcription factor (TF). Biological factors, such as chromatin state, indirect and cooperative binding, as well as experimental factors, such as antibody quality, cross-linking, and PCR biases, are known to affect the outcome of ChIP-seq experiments. However, the relative impact of these factors on inferences made from ChIP-seq data is not entirely clear. Here, via a detailed ChIP-seq simulation pipeline, ChIPulate, we assess the impact of various biological and experimental sources of variation on several outcomes of a ChIP-seq experiment, viz., the recoverability of the TF binding motif, accuracy of TF-DNA binding detection, the sensitivity of inferred TF-DNA binding strength, and number of replicates needed to confidently infer binding strength. We find that the TF motif can be recovered despite poor and non-uniform extraction and PCR amplification efficiencies. The recovery of the motif is, however, affected to a larger extent by the fraction of sites that are either cooperatively or indirectly bound. Importantly, our simulations reveal that the number of ChIP-seq replicates needed to accurately measure in vivo occupancy at high-affinity sites is larger than the recommended community standards. Our results establish statistical limits on the accuracy of inferences of protein-DNA binding from ChIP-seq and suggest that increasing the mean extraction efficiency, rather than amplification efficiency, would better improve sensitivity. The source code and instructions for running ChIPulate can be found at https://github.com/vishakad/chipulate.

DNA-binding proteins perform many key roles in biology, such as transcriptional regulation of gene expression and chromatin modification. ChIP-seq (Chromatin immunoprecipitation followed by high-throughput sequencing) is a widely used experimental technique to identify DNA-binding sites of specific proteins of interest, within cells, genome-wide. DNA fragments from genomic regions that are bound by a protein of interest, often a transcription factor (TF), are selectively extracted using specific antibodies, amplified using PCR, and sequenced. The sequences are mapped to the reference genome. Regions where many sequences map, called “peaks”, are used to infer the location of TF-bound loci (peaks), in vivo occupancy at those loci, and the sequence pattern (motif) to which the TF shows a binding affinity. But measurements of TF occupancy and motif inference are vulnerable to several biological and experimental sources of variation that are poorly understood and difficult to assess directly. Here, we simulate key steps of the ChIP-seq protocol with the aim of estimating the relative effects of various sources of variations on motif inference and binding affinity estimations. Besides providing specific insights and recommendations, we provide a general framework to simulate sequence reads in a ChIP-seq experiment, which should considerably aid in the development of software aimed at analyzing ChIP-seq data.

Detection of cooperatively bound transcription factor pairs using ChIP-seq peak intensities and expectation maximization

Transcription factors (TFs) often work cooperatively, where the binding of one TF to DNA enhances the binding affinity of a second TF to a nearby location. Such cooperative binding is important for activating gene expression from promoters and enhancers in both prokaryotic and eukaryotic cells. Existing methods to detect cooperative binding of a TF pair rely on analyzing the sequence that is bound. We propose a method that uses, instead, only ChIP-seq peak intensities and an expectation maximization (CPI-EM) algorithm. We validate our method using ChIP-seq data from cells where one of a pair of TFs under consideration has been genetically knocked out. Our algorithm relies on our observation that cooperative TF-TF binding is correlated with weak binding of one of the TFs, which we demonstrate in a variety of cell types, including E. coli, S. cerevisiae and M. musculus cells. We show that this method performs significantly better than a predictor based only on the ChIP-seq peak distance of the TFs under consideration. This suggests that peak intensities contain information that can help detect the cooperative binding of a TF pair. CPI-EM also outperforms an existing sequence-based algorithm in detecting cooperative binding. The CPI-EM algorithm is available at https://github.com/vishakad/cpi-em.

Transcriptional interference by small transcripts in proximal promoter regions

Proximal promoter regions (PPR) are heavily transcribed yielding different types of small RNAs. The act of transcription within PPRs might regulate downstream gene expression via transcriptional interference (TI). For analysis, we investigated capped and polyadenylated small RNA transcripts within PPRs of human RefSeq genes in eight different cell lines. Transcripts of our datasets overlapped with experimentally determined transcription factor binding sites (TFBS). For TFBSs intersected by these small RNA transcripts, we established negative correlation of sRNA expression levels and transcription factor (TF) DNA binding affinities; suggesting that the transcripts acted via TI. Accordingly, datasets were designated as TFbiTrs (TF-binding interfering transcripts). Expression of most TFbiTrs was restricted to certain cell lines. This facilitated the analysis of effects related to TFbiTr expression for the same RefSeq genes across cell lines. We consistently uncovered higher relative TF/DNA binding affinities and concomitantly higher expression levels for RefSeq genes in the absence of TFbiTrs. Analysis of corresponding chromatin landscapes supported these results. ChIA-PET revealed the participation of distal enhancers in TFbiTr transcription. Enhancers regulating TFbiTrs, in effect, act as repressors for corresponding downstream RefSeq genes. We demonstrate the significant impact of TI on gene expression using selected small RNA datasets.

Uncoupling evolutionary changes in DNA sequence, transcription factor occupancy and enhancer activity

Sequence variation within enhancers plays a major role in both evolution and disease, yet its functional impact on transcription factor (TF) occupancy and enhancer activity remains poorly understood. Here, we assayed the binding of five essential TFs over multiple stages of embryogenesis in two distant Drosophila species (with 1.4 substitutions per neutral site), identifying thousands of orthologous enhancers with conserved or diverged combinatorial occupancy. We used these binding signatures to dissect two properties of developmental enhancers: (1) potential TF cooperativity, using signatures of co-associations and co-divergence in TF occupancy. This revealed conserved combinatorial binding despite sequence divergence, suggesting protein-protein interactions sustain conserved collective occupancy. (2) Enhancer in-vivo activity, revealing orthologous enhancers with conserved activity despite divergence in TF occupancy. Taken together, we identify enhancers with diverged motifs yet conserved occupancy and others with diverged occupancy yet conserved activity, emphasising the need to functionally measure the effect of divergence on enhancer activity.

Crowdsourcing biomedical research: leveraging communities as innovation engines

The generation of large-scale biomedical data is creating unprecedented opportunities for basic and translational science. Typically, the data producers perform initial analyses, but it is very likely that the most informative methods may reside with other groups. Crowdsourcing the analysis of complex and massive data has emerged as a framework to find robust methodologies. When the crowdsourcing is done in the form of collaborative scientific competitions, known as Challenges, the validation of the methods is inherently addressed. Challenges also encourage open innovation, create collaborative communities to solve diverse and important biomedical problems, and foster the creation and dissemination of well-curated data repositories.

An Entropy-Based Position Projection Algorithm for Motif Discovery

Motif discovery problem is crucial for understanding the structure and function of gene expression. Over the past decades, many attempts using consensus and probability training model for motif finding are successful. However, the most existing motif discovery algorithms are still time-consuming or easily trapped in a local optimum. To overcome these shortcomings, in this paper, we propose an entropy-based position projection algorithm, called EPP, which designs a projection process to divide the dataset and explores the best local optimal solution. The experimental results on real DNA sequences, Tompa data, and ChIP-seq data show that EPP is advantageous in dealing with the motif discovery problem and outperforms current widely used algorithms.

Combinatorial Cis-regulation in Saccharomyces Species

Transcriptional control of gene expression requires interactions between the cis-regulatory elements (CREs) controlling gene promoters. We developed a sensitive computational method to identify CRE combinations with conserved spacing that does not require genome alignments. When applied to seven sensu stricto and sensu lato Saccharomyces species, 80% of the predicted interactions displayed some evidence of combinatorial transcriptional behavior in several existing datasets including: (1) chromatin immunoprecipitation data for colocalization of transcription factors, (2) gene expression data for coexpression of predicted regulatory targets, and (3) gene ontology databases for common pathway membership of predicted regulatory targets. We tested several predicted CRE interactions with chromatin immunoprecipitation experiments in a wild-type strain and strains in which a predicted cofactor was deleted. Our experiments confirmed that transcription factor (TF) occupancy at the promoters of the CRE combination target genes depends on the predicted cofactor while occupancy of other promoters is independent of the predicted cofactor. Our method has the additional advantage of identifying regulatory differences between species. By analyzing the S. cerevisiae and S. bayanus genomes, we identified differences in combinatorial cis-regulation between the species and showed that the predicted changes in gene regulation explain several of the species-specific differences seen in gene expression datasets. In some instances, the same CRE combinations appear to regulate genes involved in distinct biological processes in the two different species. The results of this research demonstrate that (1) combinatorial cis-regulation can be inferred by multi-genome analysis and (2) combinatorial cis-regulation can explain differences in gene expression between species.

Incorporating chromatin accessibility data into sequence-to-expression modeling

Prediction of gene expression levels from regulatory sequences is one of the major challenges of genomic biology today. A particularly promising approach to this problem is that taken by thermodynamics-based models that interpret an enhancer sequence in a given cellular context specified by transcription factor concentration levels and predict precise expression levels driven by that enhancer. Such models have so far not accounted for the effect of chromatin accessibility on interactions between transcription factor and DNA and consequently on gene-expression levels. Here, we extend a thermodynamics-based model of gene expression, called GEMSTAT (Gene Expression Modeling Based on Statistical Thermodynamics), to incorporate chromatin accessibility data and quantify its effect on accuracy of expression prediction. In the new model, called GEMSTAT-A, accessibility at a binding site is assumed to affect the transcription factor's binding strength at the site, whereas all other aspects are identical to the GEMSTAT model. We show that this modification results in significantly better fits in a data set of over 30 enhancers regulating spatial expression patterns in the blastoderm-stage Drosophila embryo. It is important to note that the improved fits result not from an overall elevated accessibility in active enhancers but from the variation of accessibility levels within an enhancer. With whole-genome DNA accessibility measurements becoming increasingly popular, our work demonstrates how such data may be useful for sequence-to-expression models. It also calls for future advances in modeling accessibility levels from sequence and the transregulatory context, so as to predict accurately the effect of cis and trans perturbations on gene expression.

Genome-Wide Analysis of Drosophila RBf2 Protein Highlights the Diversity of RB Family Targets and Possible Role in Regulation of Ribosome Biosynthesis

RBf2 is a recently evolved retinoblastoma family member in Drosophila that differs from RBf1, especially in the C-terminus. To investigate whether the unique features of RBf2 contribute to diverse roles in gene regulation, we performed chromatin immunoprecipitation sequencing for both RBf2 and RBf1 in embryos. A previous model for RB−E2F interactions suggested that RBf1 binds dE2F1 or dE2F2, whereas RBf2 is restricted to binding to dE2F2; however, we found that RBf2 targets approximately twice as many genes as RBf1. Highly enriched among the RBf2 targets were ribosomal protein genes. We tested the functional significance of this finding by assessing RBf activity on ribosomal protein promoters and the endogenous genes. RBf1 and RBf2 significantly repressed expression of some ribosomal protein genes, although not all bound genes showed transcriptional effects. Interestingly, many ribosomal protein genes are similarly targeted in human cells, indicating that these interactions may be relevant for control of ribosome biosynthesis and growth. We carried out bioinformatic analysis to investigate the basis for differential targeting by these two proteins and found that RBf2-specific promoters have distinct sequence motifs, suggesting unique targeting mechanisms. Association of RBf2 with these promoters appears to be independent of dE2F2/dDP, although promoters bound by both RBf1 and RBf2 require dE2F2/dDP. The presence of unique RBf2 targets suggest that evolutionary appearance of this corepressor represents the acquisition of potentially novel roles in gene regulation for the RB family.

DNA entropy reveals a significant difference in complexity between housekeeping and tissue specific gene promoters

BACKGROUND

The complexity of DNA can be quantified using estimates of entropy. Variation in DNA complexity is expected between the promoters of genes with different transcriptional mechanisms; namely housekeeping (HK) and tissue specific (TS). The former are transcribed constitutively to maintain general cellular functions, and the latter are transcribed in restricted tissue and cells types for specific molecular events. It is known that promoter features in the human genome are related to tissue specificity, but this has been difficult to quantify on a genomic scale. If entropy effectively quantifies DNA complexity, calculating the entropies of HK and TS gene promoters as profiles may reveal significant differences.

RESULTS

Entropy profiles were calculated for a total dataset of 12,003 human gene promoters and for 501 housekeeping (HK) and 587 tissue specific (TS) human gene promoters. The mean profiles show the TS promoters have a significantly lower entropy (p<2.2e-16) than HK gene promoters. The entropy distributions for the 3 datasets show that promoter entropies could be used to identify novel HK genes.

CONCLUSION

Functional features comprise DNA sequence patterns that are non-random and hence they have lower entropies. The lower entropy of TS gene promoters can be explained by a higher density of positive and negative regulatory elements, required for genes with complex spatial and temporary expression.

Integrating motif, DNA accessibility and gene expression data to build regulatory maps in an organism

Characterization of cell type specific regulatory networks and elements is a major challenge in genomics, and emerging strategies frequently employ high-throughput genome-wide assays of transcription factor (TF) to DNA binding, histone modifications or chromatin state. However, these experiments remain too difficult/expensive for many laboratories to apply comprehensively to their system of interest. Here, we explore the potential of elucidating regulatory systems in varied cell types using computational techniques that rely on only data of gene expression, low-resolution chromatin accessibility, and TF–DNA binding specificities (‘motifs’). We show that static computational motif scans overlaid with chromatin accessibility data reasonably approximate experimentally measured TF–DNA binding. We demonstrate that predicted binding profiles and expression patterns of hundreds of TFs are sufficient to identify major regulators of ∼200 spatiotemporal expression domains in the Drosophila embryo. We are then able to learn reliable statistical models of enhancer activity for over 70 expression domains and apply those models to annotate domain specific enhancers genome-wide. Throughout this work, we apply our motif and accessibility based approach to comprehensively characterize the regulatory network of fruitfly embryonic development and show that the accuracy of our computational method compares favorably to approaches that rely on data from many experimental assays.

Estimating binding properties of transcription factors from genome-wide binding profiles

The binding of transcription factors (TFs) is essential for gene expression. One important characteristic is the actual occupancy of a putative binding site in the genome. In this study, we propose an analytical model to predict genomic occupancy that incorporates the preferred target sequence of a TF in the form of a position weight matrix (PWM), DNA accessibility data (in the case of eukaryotes), the number of TF molecules expected to be bound specifically to the DNA and a parameter that modulates the specificity of the TF. Given actual occupancy data in the form of ChIP-seq profiles, we backwards inferred copy number and specificity for five Drosophila TFs during early embryonic development: Bicoid, Caudal, Giant, Hunchback and Kruppel. Our results suggest that these TFs display thousands of molecules that are specifically bound to the DNA and that whilst Bicoid and Caudal display a higher specificity, the other three TFs (Giant, Hunchback and Kruppel) display lower specificity in their binding (despite having PWMs with higher information content). This study gives further weight to earlier investigations into TF copy numbers that suggest a significant proportion of molecules are not bound specifically to the DNA.

Detecting clustering and ordering binding patterns among transcription factors via point process models

MOTIVATION

Recent development in ChIP-Seq technology has generated binding data for many transcription factors (TFs) in various cell types and cellular conditions. This opens great opportunities for studying combinatorial binding patterns among a set of TFs active in a particular cellular condition, which is a key component for understanding the interaction between TFs in gene regulation.

RESULTS

As a first step to the identification of combinatorial binding patterns, we develop statistical methods to detect clustering and ordering patterns among binding sites (BSs) of a pair of TFs. Testing procedures based on Ripley's K-function and its generalizations are developed to identify binding patterns from large collections of BSs in ChIP-Seq data. We have applied our methods to the ChIP-Seq data of 91 pairs of TFs in mouse embryonic stem cells. Our methods have detected clustering binding patterns between most TF pairs, which is consistent with the findings in the literature, and have identified significant ordering preferences, relative to the direction of target gene transcription, among the BSs of seven TFs. More interestingly, our results demonstrate that the identified clustering and ordering binding patterns between TFs are associated with the expression of the target genes. These findings provide new insights into co-regulation between TFs.

AVAILABILITY AND IMPLEMENTATION

See 'www.stat.ucla.edu/∼zhou/TFKFunctions/' for source code.

Mapping yeast transcriptional networks

The term “transcriptional network” refers to the mechanism(s) that underlies coordinated expression of genes, typically involving transcription factors (TFs) binding to the promoters of multiple genes, and individual genes controlled by multiple TFs. A multitude of studies in the last two decades have aimed to map and characterize transcriptional networks in the yeast Saccharomyces cerevisiae. We review the methodologies and accomplishments of these studies, as well as challenges we now face. For most yeast TFs, data have been collected on their sequence preferences, in vivo promoter occupancy, and gene expression profiles in deletion mutants. These systematic studies have led to the identification of new regulators of numerous cellular functions and shed light on the overall organization of yeast gene regulation. However, many yeast TFs appear to be inactive under standard laboratory growth conditions, and many of the available data were collected using techniques that have since been improved. Perhaps as a consequence, comprehensive and accurate mapping among TF sequence preferences, promoter binding, and gene expression remains an open challenge. We propose that the time is ripe for renewed systematic efforts toward a complete mapping of yeast transcriptional regulatory mechanisms.

TACO: a general-purpose tool for predicting cell-type-specific transcription factor dimers

Background

Cooperative binding of transcription factor (TF) dimers to DNA is increasingly recognized as a major contributor to binding specificity. However, it is likely that the set of known TF dimers is highly incomplete, given that they were discovered using ad hoc approaches, or through computational analyses of limited datasets.

Results

Here, we present TACO (Transcription factor Association from Complex Overrepresentation), a general-purpose standalone software tool that takes as input any genome-wide set of regulatory elements and predicts cell-type–specific TF dimers based on enrichment of motif complexes. TACO is the first tool that can accommodate motif complexes composed of overlapping motifs, a characteristic feature of many known TF dimers. Our method comprehensively outperforms existing tools when benchmarked on a reference set of 29 known dimers. We demonstrate the utility and consistency of TACO by applying it to 152 DNase-seq datasets and 94 ChIP-seq datasets.

Conclusions

Based on these results, we uncover a general principle governing the structure of TF-TF-DNA ternary complexes, namely that the flexibility of the complex is correlated with, and most likely a consequence of, inter-motif spacing.

A penalized Bayesian approach to predicting sparse protein-DNA binding landscapes

MOTIVATION

Cellular processes are controlled, directly or indirectly, by the binding of hundreds of different DNA binding factors (DBFs) to the genome. One key to deeper understanding of the cell is discovering where, when and how strongly these DBFs bind to the DNA sequence. Direct measurement of DBF binding sites (BSs; e.g. through ChIP-Chip or ChIP-Seq experiments) is expensive, noisy and not available for every DBF in every cell type. Naive and most existing computational approaches to detecting which DBFs bind in a set of genomic regions of interest often perform poorly, due to the high false discovery rates and restrictive requirements for prior knowledge.

RESULTS

We develop SparScape, a penalized Bayesian method for identifying DBFs active in the considered regions and predicting a joint probabilistic binding landscape. Using a sparsity-inducing penalization, SparScape is able to select a small subset of DBFs with enriched BSs in a set of DNA sequences from a much larger candidate set. This substantially reduces the false positives in prediction of BSs. Analysis of ChIP-Seq data in mouse embryonic stem cells and simulated data show that SparScape dramatically outperforms the naive motif scanning method and the comparable computational approaches in terms of DBF identification and BS prediction.

AVAILABILITY AND IMPLEMENTATION

SparScape is implemented in C++ with OpenMP (optional at compilation) and is freely available at 'www.stat.ucla.edu/∼zhou/Software.html' for academic use.

Physical constraints determine the logic of bacterial promoter architectures

Site-specific transcription factors (TFs) bind to their target sites on the DNA, where they regulate the rate at which genes are transcribed. Bacterial TFs undergo facilitated diffusion (a combination of 3D diffusion around and 1D random walk on the DNA) when searching for their target sites. Using computer simulations of this search process, we show that the organization of the binding sites, in conjunction with TF copy number and binding site affinity, plays an important role in determining not only the steady state of promoter occupancy, but also the order at which TFs bind. These effects can be captured by facilitated diffusion-based models, but not by standard thermodynamics. We show that the spacing of binding sites encodes complex logic, which can be derived from combinations of three basic building blocks: switches, barriers and clusters, whose response alone and in higher orders of organization we characterize in detail. Effective promoter organizations are commonly found in the E. coli genome and are highly conserved between strains. This will allow studies of gene regulation at a previously unprecedented level of detail, where our framework can create testable hypothesis of promoter logic.

Understanding variation in transcription factor binding by modeling transcription factor genome-epigenome interactions

Despite explosive growth in genomic datasets, the methods for studying epigenomic mechanisms of gene regulation remain primitive. Here we present a model-based approach to systematically analyze the epigenomic functions in modulating transcription factor-DNA binding. Based on the first principles of statistical mechanics, this model considers the interactions between epigenomic modifications and a cis-regulatory module, which contains multiple binding sites arranged in any configurations. We compiled a comprehensive epigenomic dataset in mouse embryonic stem (mES) cells, including DNA methylation (MeDIP-seq and MRE-seq), DNA hydroxymethylation (5-hmC-seq), and histone modifications (ChIP-seq). We discovered correlations of transcription factors (TFs) for specific combinations of epigenomic modifications, which we term epigenomic motifs. Epigenomic motifs explained why some TFs appeared to have different DNA binding motifs derived from in vivo (ChIP-seq) and in vitro experiments. Theoretical analyses suggested that the epigenome can modulate transcriptional noise and boost the cooperativity of weak TF binding sites. ChIP-seq data suggested that epigenomic boost of binding affinities in weak TF binding sites can function in mES cells. We showed in theory that the epigenome should suppress the TF binding differences on SNP-containing binding sites in two people. Using personal data, we identified strong associations between H3K4me2/H3K9ac and the degree of personal differences in NFκB binding in SNP-containing binding sites, which may explain why some SNPs introduce much smaller personal variations on TF binding than other SNPs. In summary, this model presents a powerful approach to analyze the functions of epigenomic modifications. This model was implemented into an open source program APEG (Affinity Prediction by Epigenome and Genome, http://systemsbio.ucsd.edu/apeg).

Predicting tissue specific transcription factor binding sites

BACKGROUND

Studies of gene regulation often utilize genome-wide predictions of transcription factor (TF) binding sites. Most existing prediction methods are based on sequence information alone, ignoring biological contexts such as developmental stages and tissue types. Experimental methods to study in vivo binding, including ChIP-chip and ChIP-seq, can only study one transcription factor in a single cell type and under a specific condition in each experiment, and therefore cannot scale to determine the full set of regulatory interactions in mammalian transcriptional regulatory networks.

RESULTS

We developed a new computational approach, PIPES, for predicting tissue-specific TF binding. PIPES integrates in vitro protein binding microarrays (PBMs), sequence conservation and tissue-specific epigenetic (DNase I hypersensitivity) information. We demonstrate that PIPES improves over existing methods on distinguishing between in vivo bound and unbound sequences using ChIP-seq data for 11 mouse TFs. In addition, our predictions are in good agreement with current knowledge of tissue-specific TF regulation.

CONCLUSIONS

We provide a systematic map of computationally predicted tissue-specific binding targets for 284 mouse TFs across 55 tissue/cell types. Such comprehensive resource is useful for researchers studying gene regulation.

Computational identification of diverse mechanisms underlying transcription factor-DNA occupancy

ChIP-based genome-wide assays of transcription factor (TF) occupancy have emerged as a powerful, high-throughput method to understand transcriptional regulation, especially on a global scale. This has led to great interest in the underlying biochemical mechanisms that direct TF-DNA binding, with the ultimate goal of computationally predicting a TF's occupancy profile in any cellular condition. In this study, we examined the influence of various potential determinants of TF-DNA binding on a much larger scale than previously undertaken. We used a thermodynamics-based model of TF-DNA binding, called “STAP,” to analyze 45 TF-ChIP data sets from Drosophila embryonic development. We built a cross-validation framework that compares a baseline model, based on the ChIP'ed (“primary”) TF's motif, to more complex models where binding by secondary TFs is hypothesized to influence the primary TF's occupancy. Candidates interacting TFs were chosen based on RNA-SEQ expression data from the time point of the ChIP experiment. We found widespread evidence of both cooperative and antagonistic effects by secondary TFs, and explicitly quantified these effects. We were able to identify multiple classes of interactions, including (1) long-range interactions between primary and secondary motifs (separated by ≤150 bp), suggestive of indirect effects such as chromatin remodeling, (2) short-range interactions with specific inter-site spacing biases, suggestive of direct physical interactions, and (3) overlapping binding sites suggesting competitive binding. Furthermore, by factoring out the previously reported strong correlation between TF occupancy and DNA accessibility, we were able to categorize the effects into those that are likely to be mediated by the secondary TF's effect on local accessibility and those that utilize accessibility-independent mechanisms. Finally, we conducted in vitro pull-down assays to test model-based predictions of short-range cooperative interactions, and found that seven of the eight TF pairs tested physically interact and that some of these interactions mediate cooperative binding to DNA.

Chromatin Immunoprecipitation (ChIP)-based genome-wide assays of transcription factor (TF) occupancy have emerged as a powerful, high throughput method to understand transcriptional regulation, especially on a global scale. Here, we utilize 45 ChIP-chip and ChIP-SEQ data sets from Drosophila to explore the underlying mechanisms of TF-DNA binding. For this, we employ a biophysically motivated computational model, in conjunction with over 300 TF motifs (binding specificities) as well as gene expression and DNA accessibility data from different developmental stages in Drosophila embryos. Our findings provide robust statistical evidence of the role played by TF-TF interactions in shaping genome-wide TF-DNA binding profiles, and thus in directing gene regulation. Our method allows us to go beyond simply recognizing the existence of such interactions, to quantifying their effects on TF occupancy. We are able to categorize the probable mechanisms of these effects as involving direct physical interactions versus accessibility-mediated indirect interactions, long-range versus short-range interactions, and cooperative versus antagonistic interactions. Our analysis reveals widespread evidence of combinatorial regulation present in recently generated ChIP data sets, and sets the stage for rich integrative models of the future that will predict cell type-specific TF occupancy values from sequence and expression data.

Comprehensive prediction in 78 human cell lines reveals rigidity and compactness of transcription factor dimers

The binding of transcription factors (TFs) to their specific motifs in genomic regulatory regions is commonly studied in isolation. However, in order to elucidate the mechanisms of transcriptional regulation, it is essential to determine which TFs bind DNA cooperatively as dimers and to infer the precise nature of these interactions. So far, only a small number of such dimeric complexes are known. Here, we present an algorithm for predicting cell-type–specific TF–TF dimerization on DNA on a large scale, using DNase I hypersensitivity data from 78 human cell lines. We represented the universe of possible TF complexes by their corresponding motif complexes, and analyzed their occurrence at cell-type–specific DNase I hypersensitive sites. Based on ∼1.4 billion tests for motif complex enrichment, we predicted 603 highly significant cell-type–specific TF dimers, the vast majority of which are novel. Our predictions included 76% (19/25) of the known dimeric complexes and showed significant overlap with an experimental database of protein–protein interactions. They were also independently supported by evolutionary conservation, as well as quantitative variation in DNase I digestion patterns. Notably, the known and predicted TF dimers were almost always highly compact and rigidly spaced, suggesting that TFs dimerize in close proximity to their partners, which results in strict constraints on the structure of the DNA-bound complex. Overall, our results indicate that chromatin openness profiles are highly predictive of cell-type–specific TF–TF interactions. Moreover, cooperative TF dimerization seems to be a widespread phenomenon, with multiple TF complexes predicted in most cell types.

Taking into account nucleosomes for predicting gene expression

The eukaryotic genome is organized in a chain of nucleosomes that consist of 145-147 bp of DNA wrapped around a histone octamer protein core. Binding of transcription factors (TF) to nucleosomal DNA is frequently impeded, which makes it a challenging task to calculate TF occupancy at a given regulatory genomic site for predicting gene expression. Here, we review methods to calculate TF binding to DNA in the presence of nucleosomes. The main theoretical problems are (i) the computation speed that is becoming a bottleneck when partial unwrapping of DNA from the nucleosome is considered, (ii) the perturbation of the binding equilibrium by the activity of ATP-dependent chromatin remodelers, which translocate nucleosomes along the DNA, and (iii) the model parameterization from high-throughput sequencing data and fluorescence microscopy experiments in living cells. We discuss strategies that address these issues to efficiently compute transcription factor binding in chromatin.

Capturing the regulatory interactions of eukaryote genomes

A key finding from early genomics research is the remarkable consistency in the number of protein-coding regions across diverse species. This has led many researchers to look to the cis-regulatory elements of genes as the fundamental influence behind evolving gene function and subsequent species diversification. Historically, since these elements are often located in vast intergenic and intronic regions of the genome, their identification has been recalcitrant. Now, with the deluge of whole-genome data from representatives of numerous eukaryotic lineages, various approaches have enabled us to begin to recognize features that characterize regulatory regions of the genome. Here we endeavour to collate these approaches in order to give an overview of the complexities involved in extrapolating regulatory signatures. The resource provided by the escalating richness of whole-genome datasets enables more sophisticated modelling of these regulatory signatures yet at the same time introduces increasing potential for noise. While we are only at the advent of making these discoveries, the next decade promises to be a very exciting and rewarding time for genome researchers.

Spatiotemporal clustering of the epigenome reveals rules of dynamic gene regulation

Spatial organization of different epigenomic marks was used to infer functions of the epigenome. It remains unclear what can be learned from the temporal changes of the epigenome. Here, we developed a probabilistic model to cluster genomic sequences based on the similarity of temporal changes of multiple epigenomic marks during a cellular differentiation process. We differentiated mouse embryonic stem (ES) cells into mesendoderm cells. At three time points during this differentiation process, we used high-throughput sequencing to measure seven histone modifications and variants—H3K4me1/2/3, H3K27ac, H3K27me3, H3K36me3, and H2A.Z; two DNA modifications—5-mC and 5-hmC; and transcribed mRNAs and noncoding RNAs (ncRNAs). Genomic sequences were clustered based on the spatiotemporal epigenomic information. These clusters not only clearly distinguished gene bodies, promoters, and enhancers, but also were predictive of bidirectional promoters, miRNA promoters, and piRNAs. This suggests specific epigenomic patterns exist on piRNA genes much earlier than germ cell development. Temporal changes of H3K4me2, unmethylated CpG, and H2A.Z were predictive of 5-hmC changes, suggesting unmethylated CpG and H3K4me2 as potential upstream signals guiding TETs to specific sequences. Several rules on combinatorial epigenomic changes and their effects on mRNA expression and ncRNA expression were derived, including a simple rule governing the relationship between 5-hmC and gene expression levels. A Sox17 enhancer containing a FOXA2 binding site and a Foxa2 enhancer containing a SOX17 binding site were identified, suggesting a positive feedback loop between the two mesendoderm transcription factors. These data illustrate the power of using epigenome dynamics to investigate regulatory functions.

Inferring direct DNA binding from ChIP-seq

Genome-wide binding data from transcription factor ChIP-seq experiments is the best source of information for inferring the relative DNA-binding affinity of these proteins in vivo. However, standard motif enrichment analysis and motif discovery approaches sometimes fail to correctly identify the binding motif for the ChIP-ed factor. To overcome this problem, we propose 'central motif enrichment analysis' (CMEA), which is based on the observation that the positional distribution of binding sites matching the direct-binding motif tends to be unimodal, well centered and maximal in the precise center of the ChIP-seq peak regions. We describe a novel visualization and statistical analysis tool--CentriMo--that identifies the region of maximum central enrichment in a set of ChIP-seq peak regions and displays the positional distributions of predicted sites. Using CentriMo for motif enrichment analysis, we provide evidence that one transcription factor (Nanog) has different binding affinity in vivo than in vitro, that another binds DNA cooperatively (E2f1), and confirm the in vivo affinity of NFIC, rescuing a difficult ChIP-seq data set. In another data set, CentriMo strongly suggests that there is no evidence of direct DNA binding by the ChIP-ed factor (Smad1). CentriMo is now part of the MEME Suite software package available at http://meme.nbcr.net. All data and output files presented here are available at: http://research.imb.uq.edu.au/t.bailey/sd/Bailey2011a.

Accurate prediction of inducible transcription factor binding intensities in vivo

DNA sequence and local chromatin landscape act jointly to determine transcription factor (TF) binding intensity profiles. To disentangle these influences, we developed an experimental approach, called protein/DNA binding followed by high-throughput sequencing (PB-seq), that allows the binding energy landscape to be characterized genome-wide in the absence of chromatin. We applied our methods to the Drosophila Heat Shock Factor (HSF), which inducibly binds a target DNA sequence element (HSE) following heat shock stress. PB-seq involves incubating sheared naked genomic DNA with recombinant HSF, partitioning the HSF-bound and HSF-free DNA, and then detecting HSF-bound DNA by high-throughput sequencing. We compared PB-seq binding profiles with ones observed in vivo by ChIP-seq and developed statistical models to predict the observed departures from idealized binding patterns based on covariates describing the local chromatin environment. We found that DNase I hypersensitivity and tetra-acetylation of H4 were the most influential covariates in predicting changes in HSF binding affinity. We also investigated the extent to which DNA accessibility, as measured by digital DNase I footprinting data, could be predicted from MNase-seq data and the ChIP-chip profiles for many histone modifications and TFs, and found GAGA element associated factor (GAF), tetra-acetylation of H4, and H4K16 acetylation to be the most predictive covariates. Lastly, we generated an unbiased model of HSF binding sequences, which revealed distinct biophysical properties of the HSF/HSE interaction and a previously unrecognized substructure within the HSE. These findings provide new insights into the interplay between the genomic sequence and the chromatin landscape in determining transcription factor binding intensity.

Quantitative models of the mechanisms that control genome-wide patterns of animal transcription factor binding

Animal transcription factors drive complex spatial and temporal patterns of gene expression during development by binding to a wide array of genomic regions. While the in vivo DNA binding landscape and in vitro DNA binding affinities of many such proteins have been characterized, our understanding of the forces that determine where, when, and the extent to which these transcription factors bind DNA in cells remains primitive. In this chapter, we describe computational thermodynamic models that predict the genome-wide DNA binding landscape of transcription factors in vivo and evaluate the contribution of biophysical determinants, such as protein-protein interactions and chromatin accessibility, on DNA occupancy. We show that predictions based only on DNA sequence and in vitro DNA affinity data achieve a mild correlation (r=0.4) with experimental measurements of in vivo DNA binding. However, by incorporating direct measurements of DNA accessibility in chromatin, it is possible to obtain much higher accuracy (r=0.6-0.9) for various transcription factors across known target genes. Thus, a combination of experimental DNA accessibility data and computational modeling of transcription factor DNA binding may be sufficient to predict the binding landscape of any animal transcription factor with reasonable accuracy.

RSAT peak-motifs: motif analysis in full-size ChIP-seq datasets

ChIP-seq is increasingly used to characterize transcription factor binding and chromatin marks at a genomic scale. Various tools are now available to extract binding motifs from peak data sets. However, most approaches are only available as command-line programs, or via a website but with size restrictions. We present peak-motifs, a computational pipeline that discovers motifs in peak sequences, compares them with databases, exports putative binding sites for visualization in the UCSC genome browser and generates an extensive report suited for both naive and expert users. It relies on time- and memory-efficient algorithms enabling the treatment of several thousand peaks within minutes. Regarding time efficiency, peak-motifs outperforms all comparable tools by several orders of magnitude. We demonstrate its accuracy by analyzing data sets ranging from 4000 to 1 28 000 peaks for 12 embryonic stem cell-specific transcription factors. In all cases, the program finds the expected motifs and returns additional motifs potentially bound by cofactors. We further apply peak-motifs to discover tissue-specific motifs in peak collections for the p300 transcriptional co-activator. To our knowledge, peak-motifs is the only tool that performs a complete motif analysis and offers a user-friendly web interface without any restriction on sequence size or number of peaks.

Evolutionary origins of transcription factor binding site clusters

Empirical studies have revealed that regulatory DNA sequences such as enhancers or promoters often harbor multiple binding sites for the same transcription factor. Such "homotypic site clustering" has been hypothesized as arising out of functional requirements of the sequences. Here, we propose an alternative explanation of this phenomenon that multisite enhancers are common because they are favored by evolutionary sampling of the genotype-phenotype landscape. To test this hypothesis, we developed a new computational framework specialized for population genetic simulations of enhancer evolution. It uses a thermodynamics-based model of enhancer function, integrating information from strong as well as weak binding sites, to determine the strength of selection. Using this framework, we found that even when simpler genotypes exist for a desired strength of regulation, relatively complex genotypes (enhancers with more sites) are more readily reached by the simulated evolutionary process. We show that there are more ways to "build" a fit genotype with many weak sites than with a few strong sites, and this is why evolution finds complex genotypes more often. Our claims are consistent with an empirical analysis of binding site content in enhancers characterized in Drosophila melanogaster and their orthologs in other Drosophila species. We also characterized a subtle but significant difference between genotypes likely to be sampled by evolution and equally fit genotypes one would obtain by uniform sampling of the fitness landscape, that is, an "evolutionary signature" in enhancer sequences. Finally, we investigated potential effects of other factors, such as rugged fitness landscapes, short local duplications, and noise characteristics of enhancers, on the emergence of homotypic site clustering. Homotypic site clustering is an important contributor to the complexity and function of cis-regulatory sequences. This work provides a simple null hypothesis for its origin, against which alternative adaptationist explanations may be evaluated, and cautions against "evolutionary mirages" present in common features of genomic sequence. The quantitative framework we develop here can be used more generally to understand how mechanisms of enhancer action influence their composition and evolution.

Bayesian analysis of high-throughput quantitative measurement of protein-DNA interactions

Transcriptional regulation depends upon the binding of transcription factor (TF) proteins to DNA in a sequence-dependent manner. Although many experimental methods address the interaction between DNA and proteins, they generally do not comprehensively and accurately assess the full binding repertoire (the complete set of sequences that might be bound with at least moderate strength). Here, we develop and evaluate through simulation an experimental approach that allows simultaneous high-throughput quantitative analysis of TF binding affinity to thousands of potential DNA ligands. Tens of thousands of putative binding targets can be mixed with a TF, and both the pre-bound and bound target pools sequenced. A hierarchical Bayesian Markov chain Monte Carlo approach determines posterior estimates for the dissociation constants, sequence-specific binding energies, and free TF concentrations. A unique feature of our approach is that dissociation constants are jointly estimated from their inferred degree of binding and from a model of binding energetics, depending on how many sequence reads are available and the explanatory power of the energy model. Careful experimental design is necessary to obtain accurate results over a wide range of dissociation constants. This approach, which we call Simultaneous Ultra high-throughput Ligand Dissociation EXperiment (SULDEX), is theoretically capable of rapid and accurate elucidation of an entire TF-binding repertoire.

Calculating transcription factor binding maps for chromatin

Current high-throughput experiments already generate enough data for retrieving the DNA sequence-dependent binding affinities of transcription factors (TF) and other chromosomal proteins throughout the complete genome. However, the reverse task of calculating binding maps in a chromatin context for a given set of concentrations and TF affinities appears to be even more challenging and computationally demanding. The problem can be addressed by considering the DNA sequence as a one-dimensional lattice with units of one or more base pairs. To calculate protein occupancies in chromatin, one needs to consider the competition of TF and histone octamers for binding sites as well as the partial unwrapping of nucleosomal DNA. Here, we consider five different classes of algorithms to compute binding maps that include the binary variable, combinatorial, sequence generating function, transfer matrix and dynamic programming approaches. The calculation time of the binary variable algorithm scales exponentially with DNA length, which limits its use to the analysis of very small genomic regions. For regulatory regions with many overlapping binding sites, potentially applicable algorithms reduce either to the transfer matrix or dynamic programming approach. In addition to the recently proposed transfer matrix formalism for TF access to the nucleosomal organized DNA, we develop here a dynamic programming algorithm that accounts for this feature. In the absence of nucleosomes, dynamic programming outperforms the transfer matrix approach, but the latter is faster when nucleosome unwrapping has to be considered. Strategies are discussed that could further facilitate calculations to allow computing genome-wide TF binding maps.

Towards an evolutionary model of transcription networks

DNA evolution models made invaluable contributions to comparative genomics, although it seemed formidable to include non-genomic features into these models. In order to build an evolutionary model of transcription networks (TNs), we had to forfeit the substitution model used in DNA evolution and to start from modeling the evolution of the regulatory relationships. We present a quantitative evolutionary model of TNs, subjecting the phylogenetic distance and the evolutionary changes of cis-regulatory sequence, gene expression and network structure to one probabilistic framework. Using the genome sequences and gene expression data from multiple species, this model can predict regulatory relationships between a transcription factor (TF) and its target genes in all species, and thus identify TN re-wiring events. Applying this model to analyze the pre-implantation development of three mammalian species, we identified the conserved and re-wired components of the TNs downstream to a set of TFs including Oct4, Gata3/4/6, cMyc and nMyc. Evolutionary events on the DNA sequence that led to turnover of TF binding sites were identified, including a birth of an Oct4 binding site by a 2nt deletion. In contrast to recent reports of large interspecies differences of TF binding sites and gene expression patterns, the interspecies difference in TF-target relationship is much smaller. The data showed increasing conservation levels from genomic sequences to TF-DNA interaction, gene expression, TN, and finally to morphology, suggesting that evolutionary changes are larger at molecular levels and smaller at functional levels. The data also showed that evolutionarily older TFs are more likely to have conserved target genes, whereas younger TFs tend to have larger re-wiring rates.

DNA evolution models made invaluable contributions to comparative genomic studies. Still lacking is an evolutionary model of transcription networks (TNs). To develop such a model, we had to forfeit the substitution model used in DNA evolution and to start from modeling the evolution of the regulatory relationships, and then subject the phylogenetic distance and the multi-species DNA sequence and gene expression data to one probabilistic framework. This model enabled us to infer the evolutionary changes of transcriptional regulatory relationships. Applying this model to analyze three yeast species, we found the anaerobic phenotype in two species was associated with the evolutionary loss of a larger cis-regulatory motif than previously thought. Analyzing three mammalian species, we found increasing conservation levels from genomic sequences to transcription factor-DNA interaction, gene expression, TN, and finally to morphology, suggesting that evolutionary changes are larger at molecular levels and smaller at functional levels. We also found that evolutionarily younger TFs are more likely to regulate different target genes in different species.

Quantitative models of the mechanisms that control genome-wide patterns of transcription factor binding during early Drosophila development

Transcription factors that drive complex patterns of gene expression during animal development bind to thousands of genomic regions, with quantitative differences in binding across bound regions mediating their activity. While we now have tools to characterize the DNA affinities of these proteins and to precisely measure their genome-wide distribution in vivo, our understanding of the forces that determine where, when, and to what extent they bind remains primitive. Here we use a thermodynamic model of transcription factor binding to evaluate the contribution of different biophysical forces to the binding of five regulators of early embryonic anterior-posterior patterning in Drosophila melanogaster. Predictions based on DNA sequence and in vitro protein-DNA affinities alone achieve a correlation of ∼0.4 with experimental measurements of in vivo binding. Incorporating cooperativity and competition among the five factors, and accounting for spatial patterning by modeling binding in every nucleus independently, had little effect on prediction accuracy. A major source of error was the prediction of binding events that do not occur in vivo, which we hypothesized reflected reduced accessibility of chromatin. To test this, we incorporated experimental measurements of genome-wide DNA accessibility into our model, effectively restricting predicted binding to regions of open chromatin. This dramatically improved our predictions to a correlation of 0.6-0.9 for various factors across known target genes. Finally, we used our model to quantify the roles of DNA sequence, accessibility, and binding competition and cooperativity. Our results show that, in regions of open chromatin, binding can be predicted almost exclusively by the sequence specificity of individual factors, with a minimal role for protein interactions. We suggest that a combination of experimentally determined chromatin accessibility data and simple computational models of transcription factor binding may be used to predict the binding landscape of any animal transcription factor with significant precision.

Assessing computational methods of cis-regulatory module prediction

Computational methods attempting to identify instances of cis-regulatory modules (CRMs) in the genome face a challenging problem of searching for potentially interacting transcription factor binding sites while knowledge of the specific interactions involved remains limited. Without a comprehensive comparison of their performance, the reliability and accuracy of these tools remains unclear. Faced with a large number of different tools that address this problem, we summarized and categorized them based on search strategy and input data requirements. Twelve representative methods were chosen and applied to predict CRMs from the Drosophila CRM database REDfly, and across the human ENCODE regions. Our results show that the optimal choice of method varies depending on species and composition of the sequences in question. When discriminating CRMs from non-coding regions, those methods considering evolutionary conservation have a stronger predictive power than methods designed to be run on a single genome. Different CRM representations and search strategies rely on different CRM properties, and different methods can complement one another. For example, some favour homotypical clusters of binding sites, while others perform best on short CRMs. Furthermore, most methods appear to be sensitive to the composition and structure of the genome to which they are applied. We analyze the principal features that distinguish the methods that performed well, identify weaknesses leading to poor performance, and provide a guide for users. We also propose key considerations for the development and evaluation of future CRM-prediction methods.

Transcriptional regulation involves multiple transcription factors binding to DNA sequences. A limited repertoire of transcription factors performs this complex regulatory step through various spatial and temporal interactions between themselves and their binding sites. These transcription factor binding interactions are clustered as distinct modules: cis-regulatory modules (CRMs). Computational methods attempting to identify instances of CRMs in the genome face a challenging problem because a majority of these interactions between transcription factors remain unknown. To investigate the reliability and accuracy of these methods, we chose twelve representative methods and applied them to predict CRMs on both the fly and human genomes. Our results show that the optimal choice of method varies depending on species and composition of the sequences in question. Different CRM representations and search strategies rely on different CRM properties, and different methods can complement one another. We provide a guide for users and key considerations for developers. We also expect that, along with new technology generating new types of genomic data, future CRM prediction methods will be able to reveal transcription binding interactions in three-dimensional space.