Improving analysis of transcription factor binding sites within ChIP-Seq data based on topological motif enrichment

Single-cell chromatin accessibility reveals malignant regulatory programs in primary human cancers

To identify cancer-associated gene regulatory changes, we generated single-cell chromatin accessibility landscapes across eight tumor types as part of The Cancer Genome Atlas. Tumor chromatin accessibility is strongly influenced by copy number alterations that can be used to identify subclones, yet underlying cis-regulatory landscapes retain cancer type-specific features. Using organ-matched healthy tissues, we identified the "nearest healthy" cell types in diverse cancers, demonstrating that the chromatin signature of basal-like-subtype breast cancer is most similar to secretory-type luminal epithelial cells. Neural network models trained to learn regulatory programs in cancer revealed enrichment of model-prioritized somatic noncoding mutations near cancer-associated genes, suggesting that dispersed, nonrecurrent, noncoding mutations in cancer are functional. Overall, these data and interpretable gene regulatory models for cancer and healthy tissue provide a framework for understanding cancer-specific gene regulation.

Genomic background sequences systematically outperform synthetic ones in de novo motif discovery for ChIP-seq data

Efficient de novo motif discovery from the results of wide-genome mapping of transcription factor binding sites (ChIP-seq) is dependent on the choice of background nucleotide sequences. The foreground sequences (ChIP-seq peaks) represent not only specific motifs of target transcription factors, but also the motifs overrepresented throughout the genome, such as simple sequence repeats. We performed a massive comparison of the 'synthetic' and 'genomic' approaches to generate background sequences for de novo motif discovery. The 'synthetic' approach shuffled nucleotides in peaks, while in the 'genomic' approach selected sequences from the reference genome randomly or only from gene promoters according to the fraction of A/T nucleotides in each sequence. We compiled the benchmark collections of ChIP-seq datasets for mouse, human and Arabidopsis, and performed de novo motif discovery. We showed that the genomic approach has both more robust detection of the known motifs of target transcription factors and more stringent exclusion of the simple sequence repeats as possible non-specific motifs. The advantage of the genomic approach over the synthetic approach was greater in plants compared to mammals. We developed the AntiNoise web service (https://denovosea.icgbio.ru/antinoise/) that implements a genomic approach to extract genomic background sequences for twelve eukaryotic genomes.

Regression convolutional neural network models implicate peripheral immune regulatory variants in the predisposition to Alzheimer's disease

Alzheimer's disease (AD) involves aggregation of amyloid β and tau, neuron loss, cognitive decline, and neuroinflammatory responses. Both resident microglia and peripheral immune cells have been associated with the immune component of AD. However, the relative contribution of resident and peripheral immune cell types to AD predisposition has not been thoroughly explored due to their similarity in gene expression and function. To study the effects of AD-associated variants on cis-regulatory elements, we train convolutional neural network (CNN) regression models that link genome sequence to cell type-specific levels of open chromatin, a proxy for regulatory element activity. We then use in silico mutagenesis of regulatory sequences to predict the relative impact of candidate variants across these cell types. We develop and apply criteria for evaluating our models and refine our models using massively parallel reporter assay (MPRA) data. Our models identify multiple AD-associated variants with a greater predicted impact in peripheral cells relative to microglia or neurons. Our results support their use as models to study the effects of AD-associated variants and even suggest that peripheral immune cells themselves may mediate a component of AD predisposition. We make our library of CNN models and predictions available as a resource for the community to study immune and neurological disorders.

TFscope: systematic analysis of the sequence features involved in the binding preferences of transcription factors

Characterizing the binding preferences of transcription factors (TFs) in different cell types and conditions is key to understand how they orchestrate gene expression. Here, we develop TFscope, a machine learning approach that identifies sequence features explaining the binding differences observed between two ChIP-seq experiments targeting either the same TF in two conditions or two TFs with similar motifs (paralogous TFs). TFscope systematically investigates differences in the core motif, nucleotide environment and co-factor motifs, and provides the contribution of each key feature in the two experiments. TFscope was applied to > 305 ChIP-seq pairs, and several examples are discussed.

Aberrant non-canonical NF-κB signalling reprograms the epigenome landscape to drive oncogenic transcriptomes in multiple myeloma

In multiple myeloma, abnormal plasma cells establish oncogenic niches within the bone marrow by engaging the NF-κB pathway to nurture their survival while they accumulate pro-proliferative mutations. Under these conditions, many cases eventually develop genetic abnormalities endowing them with constitutive NF-κB activation. Here, we find that sustained NF-κB/p52 levels resulting from such mutations favours the recruitment of enhancers beyond the normal B-cell repertoire. Furthermore, through targeted disruption of p52, we characterise how such enhancers are complicit in the formation of super-enhancers and the establishment of cis-regulatory interactions with myeloma dependencies during constitutive activation of p52. Finally, we functionally validate the pathological impact of these cis-regulatory modules on cell and tumour phenotypes using in vitro and in vivo models, confirming RGS1 as a p52-dependent myeloma driver. We conclude that the divergent epigenomic reprogramming enforced by aberrant non-canonical NF-κB signalling potentiates transcriptional programs beneficial for multiple myeloma progression.

Peak Scores Significantly Depend on the Relationships between Contextual Signals in ChIP-Seq Peaks

Chromatin immunoprecipitation followed by massively parallel DNA sequencing (ChIP-seq) is a central genome-wide method for in vivo analyses of DNA-protein interactions in various cellular conditions. Numerous studies have demonstrated the complex contextual organization of ChIP-seq peak sequences and the presence of binding sites for transcription factors in them. We assessed the dependence of the ChIP-seq peak score on the presence of different contextual signals in the peak sequences by analyzing these sequences from several ChIP-seq experiments using our fully enumerative GPU-based de novo motif discovery method, Argo_CUDA. Analysis revealed sets of significant IUPAC motifs corresponding to the binding sites of the target and partner transcription factors. For these ChIP-seq experiments, multiple regression models were constructed, demonstrating a significant dependence of the peak scores on the presence in the peak sequences of not only highly significant target motifs but also less significant motifs corresponding to the binding sites of the partner transcription factors. A significant correlation was shown between the presence of the target motifs FOXA2 and the partner motifs HNF4G, which found experimental confirmation in the scientific literature, demonstrating the important contribution of the partner transcription factors to the binding of the target transcription factor to DNA and, consequently, their important contribution to the peak score.

Relating enhancer genetic variation across mammals to complex phenotypes using machine learning

Protein-coding differences between species often fail to explain phenotypic diversity, suggesting the involvement of genomic elements that regulate gene expression such as enhancers. Identifying associations between enhancers and phenotypes is challenging because enhancer activity can be tissue-dependent and functionally conserved despite low sequence conservation. We developed the Tissue-Aware Conservation Inference Toolkit (TACIT) to associate candidate enhancers with species' phenotypes using predictions from machine learning models trained on specific tissues. Applying TACIT to associate motor cortex and parvalbumin-positive interneuron enhancers with neurological phenotypes revealed dozens of enhancer-phenotype associations, including brain size-associated enhancers that interact with genes implicated in microcephaly or macrocephaly. TACIT provides a foundation for identifying enhancers associated with the evolution of any convergently evolved phenotype in any large group of species with aligned genomes.

Genetic architecture of natural variation of cardiac performance from flies to humans

Deciphering the genetic architecture of human cardiac disorders is of fundamental importance but their underlying complexity is a major hurdle. We investigated the natural variation of cardiac performance in the sequenced inbred lines of the Drosophila Genetic Reference Panel (DGRP). Genome-wide associations studies (GWAS) identified genetic networks associated with natural variation of cardiac traits which were used to gain insights as to the molecular and cellular processes affected. Non-coding variants that we identified were used to map potential regulatory non-coding regions, which in turn were employed to predict transcription factors (TFs) binding sites. Cognate TFs, many of which themselves bear polymorphisms associated with variations of cardiac performance, were also validated by heart-specific knockdown. Additionally, we showed that the natural variations associated with variability in cardiac performance affect a set of genes overlapping those associated with average traits but through different variants in the same genes. Furthermore, we showed that phenotypic variability was also associated with natural variation of gene regulatory networks. More importantly, we documented correlations between genes associated with cardiac phenotypes in both flies and humans, which supports a conserved genetic architecture regulating adult cardiac function from arthropods to mammals. Specifically, roles for PAX9 and EGR2 in the regulation of the cardiac rhythm were established in both models, illustrating that the characteristics of natural variations in cardiac function identified in Drosophila can accelerate discovery in humans.

Motif models proposing independent and interdependent impacts of nucleotides are related to high and low affinity transcription factor binding sites in Arabidopsis

Position weight matrix (PWM) is the traditional motif model representing the transcription factor (TF) binding sites. It proposes that the positions contribute independently to TFs binding affinity, although this hypothesis does not fit the data perfectly. This explains why PWM hits are missing in a substantial fraction of ChIP-seq peaks. To study various modes of the direct binding of plant TFs, we compiled the benchmark collection of 111 ChIP-seq datasets for Arabidopsis thaliana, and applied the traditional PWM, and two alternative motif models BaMM and SiteGA, proposing the dependencies of the positions. The variation in the stringency of the recognition thresholds for the models proposed that the hits of PWM, BaMM, and SiteGA models are associated with the sites of high/medium, any, and low affinity, respectively. At the medium recognition threshold, about 60% of ChIP-seq peaks contain PWM hits consisting of conserved core consensuses, while BaMM and SiteGA provide hits for an additional 15% of peaks in which a weaker core consensus is compensated through intra-motif dependencies. The presence/absence of these dependencies in the motifs of alternative/traditional models was confirmed by the dependency logo DepLogo visualizing the position-wise partitioning of the alignments of predicted sites. We exemplify the detailed analysis of ChIP-seq profiles for plant TFs CCA1, MYC2, and SEP3. Gene ontology (GO) enrichment analysis revealed that among the three motif models, the SiteGA had the highest portions of genes with the significantly enriched GO terms among all predicted genes. We showed that both alternative motif models provide for traditional PWM greater extensions in predicted sites for TFs MYC2/SEP3 with condition/tissue specific functions, compared to those for TF CCA1 with housekeeping functions. Overall, the combined application of standard and alternative motif models is beneficial to detect various modes of the direct TF-DNA interactions in the maximal portion of ChIP-seq loci.

Addiction-Associated Genetic Variants Implicate Brain Cell Type- and Region-Specific Cis-Regulatory Elements in Addiction Neurobiology

Recent large genome-wide association studies have identified multiple confident risk loci linked to addiction-associated behavioral traits. Most genetic variants linked to addiction-associated traits lie in noncoding regions of the genome, likely disrupting cis-regulatory element (CRE) function. CREs tend to be highly cell type-specific and may contribute to the functional development of the neural circuits underlying addiction. Yet, a systematic approach for predicting the impact of risk variants on the CREs of specific cell populations is lacking. To dissect the cell types and brain regions underlying addiction-associated traits, we applied stratified linkage disequilibrium score regression to compare genome-wide association studies to genomic regions collected from human and mouse assays for open chromatin, which is associated with CRE activity. We found enrichment of addiction-associated variants in putative CREs marked by open chromatin in neuronal (NeuN+) nuclei collected from multiple prefrontal cortical areas and striatal regions known to play major roles in reward and addiction. To further dissect the cell type-specific basis of addiction-associated traits, we also identified enrichments in human orthologs of open chromatin regions of female and male mouse neuronal subtypes: cortical excitatory, D1, D2, and PV. Last, we developed machine learning models to predict mouse cell type-specific open chromatin, enabling us to further categorize human NeuN+ open chromatin regions into cortical excitatory or striatal D1 and D2 neurons and predict the functional impact of addiction-associated genetic variants. Our results suggest that different neuronal subtypes within the reward system play distinct roles in the variety of traits that contribute to addiction.SIGNIFICANCE STATEMENT We combine statistical genetic and machine learning techniques to find that the predisposition to for nicotine, alcohol, and cannabis use behaviors can be partially explained by genetic variants in conserved regulatory elements within specific brain regions and neuronal subtypes of the reward system. Our computational framework can flexibly integrate open chromatin data across species to screen for putative causal variants in a cell type- and tissue-specific manner for numerous complex traits.

Biologically relevant transfer learning improves transcription factor binding prediction

BACKGROUND

Deep learning has proven to be a powerful technique for transcription factor (TF) binding prediction but requires large training datasets. Transfer learning can reduce the amount of data required for deep learning, while improving overall model performance, compared to training a separate model for each new task.

RESULTS

We assess a transfer learning strategy for TF binding prediction consisting of a pre-training step, wherein we train a multi-task model with multiple TFs, and a fine-tuning step, wherein we initialize single-task models for individual TFs with the weights learned by the multi-task model, after which the single-task models are trained at a lower learning rate. We corroborate that transfer learning improves model performance, especially if in the pre-training step the multi-task model is trained with biologically relevant TFs. We show the effectiveness of transfer learning for TFs with ~ 500 ChIP-seq peak regions. Using model interpretation techniques, we demonstrate that the features learned in the pre-training step are refined in the fine-tuning step to resemble the binding motif of the target TF (i.e., the recipient of transfer learning in the fine-tuning step). Moreover, pre-training with biologically relevant TFs allows single-task models in the fine-tuning step to learn useful features other than the motif of the target TF.

CONCLUSIONS

Our results confirm that transfer learning is a powerful technique for TF binding prediction.

BiasAway: command-line and web server to generate nucleotide composition-matched DNA background sequences

Motivation

Accurate motif enrichment analyses depend on the choice of background DNA sequences used, which should ideally match the sequence composition of the foreground sequences. It is important to avoid false positive enrichment due to sequence biases in the genome, such as GC-bias. Therefore, relying on an appropriate set of background sequences is crucial for enrichment analysis.

Results

We developed BiasAway, a command line tool and its dedicated easy-to-use web server to generate synthetic sequences matching any k-mer nucleotide composition or select genomic DNA sequences matching the mononucleotide composition of the foreground sequences through four different models. For genomic sequences, we provide precomputed partitions of genomes from nine species with five different bin sizes to generate appropriate genomic background sequences.

Availability and implementation

BiasAway source code is freely available from Bitbucket (https://bitbucket.org/CBGR/biasaway) and can be easily installed using bioconda or pip. The web server is available at https://biasaway.uio.no and a detailed documentation is available at https://biasaway.readthedocs.io.

Supplementary information

Supplementary data are available at Bioinformatics online.

UniBind: maps of high-confidence direct TF-DNA interactions across nine species

BACKGROUND

Transcription factors (TFs) bind specifically to TF binding sites (TFBSs) at cis-regulatory regions to control transcription. It is critical to locate these TF-DNA interactions to understand transcriptional regulation. Efforts to predict bona fide TFBSs benefit from the availability of experimental data mapping DNA binding regions of TFs (chromatin immunoprecipitation followed by sequencing - ChIP-seq).

RESULTS

In this study, we processed ~ 10,000 public ChIP-seq datasets from nine species to provide high-quality TFBS predictions. After quality control, it culminated with the prediction of ~ 56 million TFBSs with experimental and computational support for direct TF-DNA interactions for 644 TFs in > 1000 cell lines and tissues. These TFBSs were used to predict > 197,000 cis-regulatory modules representing clusters of binding events in the corresponding genomes. The high-quality of the TFBSs was reinforced by their evolutionary conservation, enrichment at active cis-regulatory regions, and capacity to predict combinatorial binding of TFs. Further, we confirmed that the cell type and tissue specificity of enhancer activity was correlated with the number of TFs with binding sites predicted in these regions. All the data is provided to the community through the UniBind database that can be accessed through its web-interface ( https://unibind.uio.no/ ), a dedicated RESTful API, and as genomic tracks. Finally, we provide an enrichment tool, available as a web-service and an R package, for users to find TFs with enriched TFBSs in a set of provided genomic regions.

CONCLUSIONS

UniBind is the first resource of its kind, providing the largest collection of high-confidence direct TF-DNA interactions in nine species.

Identification of Key Tissue-Specific, Biological Processes by Integrating Enhancer Information in Maize Gene Regulatory Networks

Enhancers are key players in the spatio-temporal coordination of gene expression during numerous crucial processes, including tissue differentiation across development. Characterizing the transcription factors (TFs) and genes they connect, and the molecular functions underpinned is important to better characterize developmental processes. In plants, the recent molecular characterization of enhancers revealed their capacity to activate the expression of several target genes. Nevertheless, identifying these target genes at a genome-wide level is challenging, particularly for large-genome species, where enhancers and target genes can be hundreds of kilobases away. Therefore, the contribution of enhancers to plant regulatory networks remains poorly understood. Here, we investigate the enhancer-driven regulatory network of two maize tissues at different stages: leaves at seedling stage (V2-IST) and husks (bracts) at flowering. Using systems biology, we integrate genomic, epigenomic, and transcriptomic data to model the regulatory relationships between TFs and their potential target genes, and identify regulatory modules specific to husk and V2-IST. We show that leaves at the V2-IST stage are characterized by the response to hormones and macromolecules biogenesis and assembly, which are regulated by the BBR/BPC and AP2/ERF TF families, respectively. In contrast, husks are characterized by cell wall modification and response to abiotic stresses, which are, respectively, orchestrated by the C2C2/DOF and AP2/EREB families. Analysis of the corresponding enhancer sequences reveals that two different transposable element families (TIR transposon Mutator and MITE Pif/Harbinger) have shaped part of the regulatory network in each tissue, and that MITEs have provided potential new TF binding sites involved in husk tissue-specificity.

MEIRLOP: improving score-based motif enrichment by incorporating sequence bias covariates

BACKGROUND

Motif enrichment analysis (MEA) identifies over-represented transcription factor binding (TF) motifs in the DNA sequence of regulatory regions, enabling researchers to infer which transcription factors can regulate transcriptional response to a stimulus, or identify sequence features found near a target protein in a ChIP-seq experiment. Score-based MEA determines motifs enriched in regions exhibiting extreme differences in regulatory activity, but existing methods do not control for biases in GC content or dinucleotide composition. This lack of control for sequence bias, such as those often found in CpG islands, can obscure the enrichment of biologically relevant motifs.

RESULTS

We developed Motif Enrichment In Ranked Lists of Peaks (MEIRLOP), a novel MEA method that determines enrichment of TF binding motifs in a list of scored regulatory regions, while controlling for sequence bias. In this study, we compare MEIRLOP against other MEA methods in identifying binding motifs found enriched in differentially active regulatory regions after interferon-beta stimulus, finding that using logistic regression and covariates improves the ability to call enrichment of ISGF3 binding motifs from differential acetylation ChIP-seq data compared to other methods. Our method achieves similar or better performance compared to other methods when quantifying the enrichment of TF binding motifs from ENCODE TF ChIP-seq datasets. We also demonstrate how MEIRLOP is broadly applicable to the analysis of numerous types of NGS assays and experimental designs.

CONCLUSIONS

Our results demonstrate the importance of controlling for sequence bias when accurately identifying enriched DNA sequence motifs using score-based MEA. MEIRLOP is available for download from https://github.com/npdeloss/meirlop under the MIT license.

Occupancy maps of 208 chromatin-associated proteins in one human cell type

Transcription factors are DNA-binding proteins that have key roles in gene regulation^1,2. Genome-wide occupancy maps of transcriptional regulators are important for understanding gene regulation and its effects on diverse biological processes^3–6. However, only a minority of the more than 1,600 transcription factors encoded in the human genome has been assayed. Here we present, as part of the ENCODE (Encyclopedia of DNA Elements) project, data and analyses from chromatin immunoprecipitation followed by high-throughput sequencing (ChIP–seq) experiments using the human HepG2 cell line for 208 chromatin-associated proteins (CAPs). These comprise 171 transcription factors and 37 transcriptional cofactors and chromatin regulator proteins, and represent nearly one-quarter of CAPs expressed in HepG2 cells. The binding profiles of these CAPs form major groups associated predominantly with promoters or enhancers, or with both. We confirm and expand the current catalogue of DNA sequence motifs for transcription factors, and describe motifs that correspond to other transcription factors that are co-enriched with the primary ChIP target. For example, FOX family motifs are enriched in ChIP–seq peaks of 37 other CAPs. We show that motif content and occupancy patterns can distinguish between promoters and enhancers. This catalogue reveals high-occupancy target regions at which many CAPs associate, although each contains motifs for only a minority of the numerous associated transcription factors. These analyses provide a more complete overview of the gene regulatory networks that define this cell type, and demonstrate the usefulness of the large-scale production efforts of the ENCODE Consortium.

ChIP–seq and CETCh–seq data are used to analyse binding maps for 208 transcription factors and other chromatin-associated proteins in a single human cell type, providing a comprehensive catalogue of the transcription factor landscape and gene regulatory networks in these cells.

Mechanistic insights into transcription factor cooperativity and its impact on protein-phenotype interactions

Recent high-throughput transcription factor (TF) binding assays revealed that TF cooperativity is a widespread phenomenon. However, a global mechanistic and functional understanding of TF cooperativity is still lacking. To address this, here we introduce a statistical learning framework that provides structural insight into TF cooperativity and its functional consequences based on next generation sequencing data. We identify DNA shape as driver for cooperativity, with a particularly strong effect for Forkhead-Ets pairs. Follow-up experiments reveal a local shape preference at the Ets-DNA-Forkhead interface and decreased cooperativity upon loss of the interaction. Additionally, we discover many functional associations for cooperatively bound TFs. Examination of the link between FOXO1:ETV6 and lymphomas reveals that their joint expression levels improve patient clinical outcome stratification. Altogether, our results demonstrate that inter-family cooperative TF binding is driven by position-specific DNA readout mechanisms, which provides an additional regulatory layer for downstream biological functions.

Although transcription factor (TF) cooperativity is widespread, a global mechanistic understanding of the role of TF cooperativity is still lacking. Here the authors introduce a statistical learning framework that provides structural insight into TF cooperativity and its functional consequences based on next generation sequencing data and provide mechanistic insights into TF cooperativity and its impact on protein-phenotype interactions.

Diverse families of transposable elements affect the transcriptional regulation of stress-response genes in Drosophila melanogaster

Although transposable elements are an important source of regulatory variation, their genome-wide contribution to the transcriptional regulation of stress-response genes has not been studied yet. Stress is a major aspect of natural selection in the wild, leading to changes in the transcriptional regulation of a variety of genes that are often triggered by one or a few transcription factors. In this work, we take advantage of the wealth of information available for Drosophila melanogaster and humans to analyze the role of transposable elements in six stress regulatory networks: immune, hypoxia, oxidative, xenobiotic, heat shock, and heavy metal. We found that transposable elements were enriched for caudal, dorsal, HSF, and tango binding sites in D. melanogaster and for NFE2L2 binding sites in humans. Taking into account the D. melanogaster population frequencies of transposable elements with predicted binding motifs and/or binding sites, we showed that those containing three or more binding motifs/sites are more likely to be functional. For a representative subset of these TEs, we performed in vivo transgenic reporter assays in different stress conditions. Overall, our results showed that TEs are relevant contributors to the transcriptional regulation of stress-response genes.

The Mycobacterial LexA/RecA-Independent DNA Damage Response Is Controlled by PafBC and the Pup-Proteasome System

Mycobacteria exhibit two DNA damage response pathways: the LexA/RecA-dependent SOS response and a LexA/RecA-independent pathway. Using a combination of transcriptomics and genome-wide binding site analysis, we demonstrate that PafBC (proteasome accessory factor B and C), encoded in the Pup-proteasome system (PPS) gene locus, is the transcriptional regulator of the predominant LexA/RecA-independent pathway. Comparison of the resulting PafBC regulon with the DNA damage response of Mycobacterium smegmatis reveals that the majority of induced DNA repair genes are upregulated by PafBC. We further demonstrate that RecA, a member of the PafBC regulon and principal regulator of the SOS response, is degraded by the PPS when DNA damage stress has been overcome. Our results suggest a model for the regulation of the mycobacterial DNA damage response that employs the concerted action of PafBC as master transcriptional activator and the PPS for removal of DNA repair proteins to maintain a temporally controlled stress response.

A map of direct TF-DNA interactions in the human genome

Chromatin immunoprecipitation followed by sequencing (ChIP-seq) is the most popular assay to identify genomic regions, called ChIP-seq peaks, that are bound in vivo by transcription factors (TFs). These regions are derived from direct TF-DNA interactions, indirect binding of the TF to the DNA (through a co-binding partner), nonspecific binding to the DNA, and noise/bias/artifacts. Delineating the bona fide direct TF-DNA interactions within the ChIP-seq peaks remains challenging. We developed a dedicated software, ChIP-eat, that combines computational TF binding models and ChIP-seq peaks to automatically predict direct TF-DNA interactions. Our work culminated with predicted interactions covering >4% of the human genome, obtained by uniformly processing 1983 ChIP-seq peak data sets from the ReMap database for 232 unique TFs. The predictions were a posteriori assessed using protein binding microarray and ChIP-exo data, and were predominantly found in high quality ChIP-seq peaks. The set of predicted direct TF-DNA interactions suggested that high-occupancy target regions are likely not derived from direct binding of the TFs to the DNA. Our predictions derived co-binding TFs supported by protein-protein interaction data and defined cis-regulatory modules enriched for disease- and trait-associated SNPs. We provide this collection of direct TF-DNA interactions and cis-regulatory modules through the UniBind web-interface (http://unibind.uio.no).

BiFET: sequencing Bias-free transcription factor Footprint Enrichment Test

Transcription factor (TF) footprinting uncovers putative protein–DNA binding via combined analyses of chromatin accessibility patterns and their underlying TF sequence motifs. TF footprints are frequently used to identify TFs that regulate activities of cell/condition-specific genomic regions (target loci) in comparison to control regions (background loci) using standard enrichment tests. However, there is a strong association between the chromatin accessibility level and the GC content of a locus and the number and types of TF footprints that can be detected at this site. Traditional enrichment tests (e.g. hypergeometric) do not account for this bias and inflate false positive associations. Therefore, we developed a novel post-processing method, Bias-free Footprint Enrichment Test (BiFET), that corrects for the biases arising from the differences in chromatin accessibility levels and GC contents between target and background loci in footprint enrichment analyses. We applied BiFET on TF footprint calls obtained from EndoC-βH1 ATAC-seq samples using three different algorithms (CENTIPEDE, HINT-BC and PIQ) and showed BiFET’s ability to increase power and reduce false positive rate when compared to hypergeometric test. Furthermore, we used BiFET to study TF footprints from human PBMC and pancreatic islet ATAC-seq samples to show its utility to identify putative TFs associated with cell-type-specific loci.

Human Enhancers Harboring Specific Sequence Composition, Activity, and Genome Organization Are Linked to the Immune Response

The FANTOM5 consortium recently characterized 65,423 human enhancers from 1829 cell and tissue samples using the Cap Analysis of Gene Expression technology. We showed that the guanine and cytosine content at enhancer regions distinguishes two classes of enhancers harboring distinct DNA structural properties at flanking regions. A functional analysis of their predicted gene targets highlighted one class of enhancers as significantly enriched for associations with immune response genes. Moreover, these enhancers were specifically enriched for regulatory motifs recognized by transcription factors involved in immune response. We observed that enhancers enriched for links to immune response genes were more cell-type specific, preferentially activated upon bacterial infection, and with specific response activity. Looking at chromatin capture data, we found that the two classes of enhancers were lying in distinct topologically associating domains and chromatin loops. Our results suggest that specific nucleotide compositions encode for classes of enhancers that are functionally distinct and specifically organized in the human genome.

DeFine: deep convolutional neural networks accurately quantify intensities of transcription factor-DNA binding and facilitate evaluation of functional non-coding variants

The complex system of gene expression is regulated by the cell type-specific binding of transcription factors (TFs) to regulatory elements. Identifying variants that disrupt TF binding and lead to human diseases remains a great challenge. To address this, we implement sequence-based deep learning models that accurately predict the TF binding intensities to given DNA sequences. In addition to accurately classifying TF-DNA binding or unbinding, our models are capable of accurately predicting real-valued TF binding intensities by leveraging large-scale TF ChIP-seq data. The changes in the TF binding intensities between the altered sequence and the reference sequence reflect the degree of functional impact for the variant. This enables us to develop the tool DeFine (Deep learning based Functional impact of non-coding variants evaluator, http://define.cbi.pku.edu.cn) with improved performance for assessing the functional impact of non-coding variants including SNPs and indels. DeFine accurately identifies the causal functional non-coding variants from disease-associated variants in GWAS. DeFine is an effective and easy-to-use tool that facilities systematic prioritization of functional non-coding variants.

Widespread activation of antisense transcription of the host genome during herpes simplex virus 1 infection

Background

Herpesviruses can infect a wide range of animal species. Herpes simplex virus 1 (HSV-1) is one of the eight herpesviruses that can infect humans and is prevalent worldwide. Herpesviruses have evolved multiple ways to adapt the infected cells to their needs, but knowledge about these transcriptional and post-transcriptional modifications is sparse.

Results

Here, we show that HSV-1 induces the expression of about 1000 antisense transcripts from the human host cell genome. A subset of these is also activated by the closely related varicella zoster virus. Antisense transcripts originate either at gene promoters or within the gene body, and they show different susceptibility to the inhibition of early and immediate early viral gene expression. Overexpression of the major viral transcription factor ICP4 is sufficient to turn on a subset of antisense transcripts. Histone marks around transcription start sites of HSV-1-induced and constitutively transcribed antisense transcripts are highly similar, indicating that the genetic loci are already poised to transcribe these novel RNAs. Furthermore, an antisense transcript overlapping with the BBC3 gene (also known as PUMA) transcriptionally silences this potent inducer of apoptosis in cis.

Conclusions

We show for the first time that a virus induces widespread antisense transcription of the host cell genome. We provide evidence that HSV-1 uses this to downregulate a strong inducer of apoptosis. Our findings open new perspectives on global and specific alterations of host cell transcription by viruses.

Electronic supplementary material

The online version of this article (doi:10.1186/s13059-017-1329-5) contains supplementary material, which is available to authorized users.

Identification of Human Lineage-Specific Transcriptional Coregulators Enabled by a Glossary of Binding Modules and Tunable Genomic Backgrounds

Transcription factors (TFs) control cellular processes by binding specific DNA motifs to modulate gene expression. Motif enrichment analysis of regulatory regions can identify direct and indirect TF binding sites. Here, we created a glossary of 108 non-redundant TF-8mer "modules" of shared specificity for 671 metazoan TFs from publicly available and new universal protein binding microarray data. Analysis of 239 ENCODE TF chromatin immunoprecipitation sequencing datasets and associated RNA sequencing profiles suggest the 8mer modules are more precise than position weight matrices in identifying indirect binding motifs and their associated tethering TFs. We also developed GENRE (genomically equivalent negative regions), a tunable tool for construction of matched genomic background sequences for analysis of regulatory regions. GENRE outperformed four state-of-the-art approaches to background sequence construction. We used our TF-8mer glossary and GENRE in the analysis of the indirect binding motifs for the co-occurrence of tethering factors, suggesting novel TF-TF interactions. We anticipate that these tools will aid in elucidating tissue-specific gene-regulatory programs.

Evaluating tools for transcription factor binding site prediction

Background

Binding of transcription factors to transcription factor binding sites (TFBSs) is key to the mediation of transcriptional regulation. Information on experimentally validated functional TFBSs is limited and consequently there is a need for accurate prediction of TFBSs for gene annotation and in applications such as evaluating the effects of single nucleotide variations in causing disease. TFBSs are generally recognized by scanning a position weight matrix (PWM) against DNA using one of a number of available computer programs. Thus we set out to evaluate the best tools that can be used locally (and are therefore suitable for large-scale analyses) for creating PWMs from high-throughput ChIP-Seq data and for scanning them against DNA.

Results

We evaluated a set of de novo motif discovery tools that could be downloaded and installed locally using ENCODE-ChIP-Seq data and showed that rGADEM was the best-performing tool. TFBS prediction tools used to scan PWMs against DNA fall into two classes — those that predict individual TFBSs and those that identify clusters. Our evaluation showed that FIMO and MCAST performed best respectively.

Conclusions

Selection of the best-performing tools for generating PWMs from ChIP-Seq data and for scanning PWMs against DNA has the potential to improve prediction of precise transcription factor binding sites within regions identified by ChIP-Seq experiments for gene finding, understanding regulation and in evaluating the effects of single nucleotide variations in causing disease.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-016-1298-9) contains supplementary material, which is available to authorized users.

DNA Shape Features Improve Transcription Factor Binding Site Predictions In Vivo

Interactions of transcription factors (TFs) with DNA comprise a complex interplay between base-specific amino acid contacts and readout of DNA structure. Recent studies have highlighted the complementarity of DNA sequence and shape in modeling TF binding in vitro. Here, we have provided a comprehensive evaluation of in vivo datasets to assess the predictive power obtained by augmenting various DNA sequence-based models of TF binding sites (TFBSs) with DNA shape features (helix twist, minor groove width, propeller twist, and roll). Results from 400 human ChIP-seq datasets for 76 TFs show that combining DNA shape features with position-specific scoring matrix (PSSM) scores improves TFBS predictions. Improvement has also been observed using TF flexible models and a machine-learning approach using a binary encoding of nucleotides in lieu of PSSMs. Incorporating DNA shape information is most beneficial for E2F and MADS-domain TF families. Our findings indicate that incorporating DNA sequence and shape information benefits the modeling of TF binding under complex in vivo conditions.

Evaluating the impact of single nucleotide variants on transcription factor binding

Diseases and phenotypes caused by disrupted transcription factor (TF) binding are being identified, but progress is hampered by our limited capacity to predict such functional alterations. Improving predictions may be dependent on expanding the set of bona fide TF binding alterations. Allele-specific binding (ASB) events, where TFs preferentially bind to one of the two alleles at heterozygous sites, reveal the impact of sequence variations in altered TF binding. Here, we present the largest ASB compilation to our knowledge, 10 765 ASB events retrieved from 45 ENCODE ChIP-Seq data sets. Our analysis showed that ASB events were frequently associated with motif alterations of the ChIP'ed TF and potential partner TFs, allelic difference of DNase I hypersensitivity and allelic difference of histone modifications. For TF dimers bound symmetrically to DNA, ASB data revealed that central positions of the TF binding motifs were disproportionately important for binding. Lastly, the impact of variation on TF binding was predicted by a classification model incorporating all the investigated features of ASB events. Classification models using only DNase I hypersensitivity and sequence data exhibited predictive accuracy approaching the models with substantially more features. Taken together, the combination of ASB data and the classification model represents an important step toward elucidating regulatory variants across the human genome.

Differences in the Early Development of Human and Mouse Embryonic Stem Cells

We performed a systematic analysis of gene expression features in early (10–21 days) development of human vs mouse embryonic cells (hESCs vs mESCs). Many development features were found to be conserved, and a majority of differentially regulated genes have similar expression change in both organisms. The similarity is especially evident, when gene expression profiles are clustered together and properties of clustered groups of genes are compared. First 10 days of mESC development match the features of hESC development within 21 days, in accordance with the differences in population doubling time in human and mouse ESCs. At the same time, several important differences are seen. There is a clear difference in initial expression change of transcription factors and stimulus responsive genes, which may be caused by the difference in experimental procedures. However, we also found that some biological processes develop differently; this can clearly be shown, for example, for neuron and sensory organ development. Some groups of genes show peaks of the expression levels during the development and these peaks cannot be claimed to happen at the same time points in the two organisms, as well as for the same groups of (orthologous) genes. We also detected a larger number of upregulated genes during development of mESCs as compared to hESCs. The differences were quantified by comparing promoters of related genes. Most of gene groups behave similarly and have similar transcription factor (TF) binding sites on their promoters. A few groups of genes have similar promoters, but are expressed differently in two species. Interestingly, there are groups of genes expressed similarly, although they have different promoters, which can be shown by comparing their TF binding sites. Namely, a large group of similarly expressed cell cycle-related genes is found to have discrepant TF binding properties in mouse vs human.

Optimally choosing PWM motif databases and sequence scanning approaches based on ChIP-seq data

BACKGROUND

For many years now, binding preferences of Transcription Factors have been described by so called motifs, usually mathematically defined by position weight matrices or similar models, for the purpose of predicting potential binding sites. However, despite the availability of thousands of motif models in public and commercial databases, a researcher who wants to use them is left with many competing methods of identifying potential binding sites in a genome of interest and there is little published information regarding the optimality of different choices. Thanks to the availability of large number of different motif models as well as a number of experimental datasets describing actual binding of TFs in hundreds of TF-ChIP-seq pairs, we set out to perform a comprehensive analysis of this matter.

RESULTS

We focus on the task of identifying potential transcription factor binding sites in the human genome. Firstly, we provide a comprehensive comparison of the coverage and quality of models available in different databases, showing that the public databases have comparable TFs coverage and better motif performance than commercial databases. Secondly, we compare different motif scanners showing that, regardless of the database used, the tools developed by the scientific community outperform the commercial tools. Thirdly, we calculate for each motif a detection threshold optimizing the accuracy of prediction. Finally, we provide an in-depth comparison of different methods of choosing thresholds for all motifs a priori. Surprisingly, we show that selecting a common false-positive rate gives results that are the least biased by the information content of the motif and therefore most uniformly accurate.

CONCLUSION

We provide a guide for researchers working with transcription factor motifs. It is supplemented with detailed results of the analysis and the benchmark datasets at http://bioputer.mimuw.edu.pl/papers/motifs/ .

Non-targeted transcription factors motifs are a systemic component of ChIP-seq datasets

Background

The global effort to annotate the non-coding portion of the human genome relies heavily on chromatin immunoprecipitation data generated with high-throughput DNA sequencing (ChIP-seq). ChIP-seq is generally successful in detailing the segments of the genome bound by the immunoprecipitated transcription factor (TF), however almost all datasets contain genomic regions devoid of the canonical motif for the TF. It remains to be determined if these regions are related to the immunoprecipitated TF or whether, despite the use of controls, there is a portion of peaks that can be attributed to other causes.

Results

Analyses across hundreds of ChIP-seq datasets generated for sequence-specific DNA binding TFs reveal a small set of TF binding profiles for which predicted TF binding site motifs are repeatedly observed to be significantly enriched. Grouping related binding profiles, the set includes: CTCF-like, ETS-like, JUN-like, and THAP11 profiles. These frequently enriched profiles are termed ‘zingers’ to highlight their unanticipated enrichment in datasets for which they were not the targeted TF, and their potential impact on the interpretation and analysis of TF ChIP-seq data. Peaks with zinger motifs and lacking the ChIPped TF’s motif are observed to compose up to 45% of a ChIP-seq dataset. There is substantial overlap of zinger motif containing regions between diverse TF datasets, suggesting a mechanism that is not TF-specific for the recovery of these regions.

Conclusions

Based on the zinger regions proximity to cohesin-bound segments, a loading station model is proposed. Further study of zingers will advance understanding of gene regulation.

Electronic supplementary material

The online version of this article (doi:10.1186/s13059-014-0412-4) contains supplementary material, which is available to authorized users.