Predicting enhancer activity and variant impact using gkm-SVM

Gapped-kmer sequence modeling robustly identifies regulatory vocabularies and distal enhancers conserved between evolutionarily distant mammals

Gene regulatory elements drive complex biological phenomena and their mutations are associated with common human diseases. The impacts of human regulatory variants are often tested using model organisms such as mice. However, mapping human enhancers to conserved elements in mice remains a challenge, due to both rapid enhancer evolution and limitations of current computational methods. We analyze distal enhancers across 45 matched human/mouse cell/tissue pairs from a comprehensive dataset of DNase-seq experiments, and show that while cell-specific regulatory vocabulary is conserved, enhancers evolve more rapidly than promoters and CTCF binding sites. Enhancer conservation rates vary across cell types, in part explainable by tissue specific transposable element activity. We present an improved genome alignment algorithm using gapped-kmer features, called gkm-align, and make genome wide predictions for 1,401,803 orthologous regulatory elements. We show that gkm-align discovers 23,660 novel human/mouse conserved enhancers missed by previous algorithms, with strong evidence of conserved functional activity.

Allele-specific binding variants causing ChIP-seq peak height of histone modification are not enriched in expression QTL annotations

BACKGROUND

Genome sequence variants affecting complex traits (quantitative trait loci, QTL) are enriched in functional regions of the genome, such as those marked by certain histone modifications. These variants are believed to influence gene expression. However, due to the linkage disequilibrium among nearby variants, pinpointing the precise location of QTL is challenging. We aimed to identify allele-specific binding (ASB) QTL (asbQTL) that cause variation in the level of histone modification, as measured by the height of peaks assayed by ChIP-seq (chromatin immunoprecipitation sequencing). We identified DNA sequences that predict the difference between alleles in ChIP-seq peak height in H3K4me3 and H3K27ac histone modifications in the mammary glands of cows.

RESULTS

We used a gapped k-mer support vector machine, a novel best linear unbiased prediction model, and a multiple linear regression model that combines the other two approaches to predict variant impacts on peak height. For each method, a subset of 1000 sites with the highest magnitude of predicted ASB was considered as candidate asbQTL. The accuracy of this prediction was measured by the proportion where the predicted direction matched the observed direction. Prediction accuracy ranged between 0.59 and 0.74, suggesting that these 1000 sites are enriched for asbQTL. Using independent data, we investigated functional enrichment in the candidate asbQTL set and three control groups, including non-causal ASB sites, non-ASB variants under a peak, and SNPs (single nucleotide polymorphisms) not under a peak. For H3K4me3, a higher proportion of the candidate asbQTL were confirmed as ASB when compared to the non-causal ASB sites (P < 0.01). However, these candidate asbQTL did not enrich for the other annotations, including expression QTL (eQTL), allele-specific expression QTL (aseQTL) and sites conserved across mammals (P > 0.05).

CONCLUSIONS

We identified putatively causal sites for asbQTL using the DNA sequence surrounding these sites. Our results suggest that many sites influencing histone modifications may not directly affect gene expression. However, it is important to acknowledge that distinguishing between putative causal ASB sites and other non-causal ASB sites in high linkage disequilibrium with the causal sites regarding their impact on gene expression may be challenging due to limitations in statistical power.

Machine learning identifies activation of RUNX/AP-1 as drivers of mesenchymal and fibrotic regulatory programs in gastric cancer

Gastric cancer (GC) is the fifth most common cancer worldwide and is a heterogeneous disease. Among GC subtypes, the mesenchymal phenotype (Mes-like) is more invasive than the epithelial phenotype (Epi-like). Although gene expression of the epithelial-to-mesenchymal transition (EMT) has been studied, the regulatory landscape shaping this process is not fully understood. Here we use ATAC-seq and RNA-seq data from a compendium of GC cell lines and primary tumors to detect drivers of regulatory state changes and their transcriptional responses. Using the ATAC-seq data, we developed a machine learning approach to determine the transcription factors (TFs) regulating the subtypes of GC. We identified TFs driving the mesenchymal (RUNX2, ZEB1, SNAI2, AP-1 dimer) and the epithelial (GATA4, GATA6, KLF5, HNF4A, FOXA2, GRHL2) states in GC. We identified DNA copy number alterations associated with dysregulation of these TFs, specifically deletion of GATA4 and amplification of MAPK9 Comparisons with bulk and single-cell RNA-seq data sets identified activation toward fibroblast-like epigenomic and expression signatures in Mes-like GC. The activation of this mesenchymal fibrotic program is associated with differentially accessible DNA cis-regulatory elements flanking upregulated mesenchymal genes. These findings establish a map of TF activity in GC and highlight the role of copy number driven alterations in shaping epigenomic regulatory programs as potential drivers of GC heterogeneity and progression.

CAGI, the Critical Assessment of Genome Interpretation, establishes progress and prospects for computational genetic variant interpretation methods

BACKGROUND

The Critical Assessment of Genome Interpretation (CAGI) aims to advance the state-of-the-art for computational prediction of genetic variant impact, particularly where relevant to disease. The five complete editions of the CAGI community experiment comprised 50 challenges, in which participants made blind predictions of phenotypes from genetic data, and these were evaluated by independent assessors.

RESULTS

Performance was particularly strong for clinical pathogenic variants, including some difficult-to-diagnose cases, and extends to interpretation of cancer-related variants. Missense variant interpretation methods were able to estimate biochemical effects with increasing accuracy. Assessment of methods for regulatory variants and complex trait disease risk was less definitive and indicates performance potentially suitable for auxiliary use in the clinic.

CONCLUSIONS

Results show that while current methods are imperfect, they have major utility for research and clinical applications. Emerging methods and increasingly large, robust datasets for training and assessment promise further progress ahead.

Comparative analysis of models in predicting the effects of SNPs on TF-DNA binding using large-scale in vitro and in vivo data

Non-coding variants associated with complex traits can alter the motifs of transcription factor (TF)-deoxyribonucleic acid binding. Although many computational models have been developed to predict the effects of non-coding variants on TF binding, their predictive power lacks systematic evaluation. Here we have evaluated 14 different models built on position weight matrices (PWMs), support vector machines, ordinary least squares and deep neural networks (DNNs), using large-scale in vitro (i.e. SNP-SELEX) and in vivo (i.e. allele-specific binding, ASB) TF binding data. Our results show that the accuracy of each model in predicting SNP effects in vitro significantly exceeds that achieved in vivo. For in vitro variant impact prediction, kmer/gkm-based machine learning methods (deltaSVM_HT-SELEX, QBiC-Pred) trained on in vitro datasets exhibit the best performance. For in vivo ASB variant prediction, DNN-based multitask models (DeepSEA, Sei, Enformer) trained on the ChIP-seq dataset exhibit relatively superior performance. Among the PWM-based methods, tRap demonstrates better performance in both in vitro and in vivo evaluations. In addition, we find that TF classes such as basic leucine zipper factors could be predicted more accurately, whereas those such as C2H2 zinc finger factors are predicted less accurately, aligning with the evolutionary conservation of these TF classes. We also underscore the significance of non-sequence factors such as cis-regulatory element type, TF expression, interactions and post-translational modifications in influencing the in vivo predictive performance of TFs. Our research provides valuable insights into selecting prioritization methods for non-coding variants and further optimizing such models.

Deep molecular learning of transcriptional control of a synthetic CRE enhancer and its variants

Massively parallel reporter assay measures transcriptional activities of various cis-regulatory modules (CRMs) in a single experiment. We developed a thermodynamic computational model framework that calculates quantitative levels of gene expression directly from regulatory DNA sequences. Using the framework, we investigated the molecular mechanisms of cis-regulatory mutations of a synthetic enhancer that cause abnormal gene expression. We found that, in a human cell line, competitive binding between family transcription factors (TFs) with slightly different binding preferences significantly increases the accuracy of recapitulating the transcriptional effects of thousands of single- or multi-mutations. We also discovered that even if various harmful mutations occurred in an activator binding site, CRM could stably maintain or even increase gene expression through a certain form of competitive binding between family TFs. These findings enhance understanding the effect of SNPs and indels on CRMs and would help building robust custom-designed CRMs for biologics production and gene therapy.

The Impact of Genomic Variation on Function (IGVF) Consortium

Our genomes influence nearly every aspect of human biology from molecular and cellular functions to phenotypes in health and disease. Human genetics studies have now associated hundreds of thousands of differences in our DNA sequence ("genomic variation") with disease risk and other phenotypes, many of which could reveal novel mechanisms of human biology and uncover the basis of genetic predispositions to diseases, thereby guiding the development of new diagnostics and therapeutics. Yet, understanding how genomic variation alters genome function to influence phenotype has proven challenging. To unlock these insights, we need a systematic and comprehensive catalog of genome function and the molecular and cellular effects of genomic variants. Toward this goal, the Impact of Genomic Variation on Function (IGVF) Consortium will combine approaches in single-cell mapping, genomic perturbations, and predictive modeling to investigate the relationships among genomic variation, genome function, and phenotypes. Through systematic comparisons and benchmarking of experimental and computational methods, we aim to create maps across hundreds of cell types and states describing how coding variants alter protein activity, how noncoding variants change the regulation of gene expression, and how both coding and noncoding variants may connect through gene regulatory and protein interaction networks. These experimental data, computational predictions, and accompanying standards and pipelines will be integrated into an open resource that will catalyze community efforts to explore genome function and the impact of genetic variation on human biology and disease across populations.

Integrating reduced amino acid composition into PSSM for improving copper ion-binding protein prediction

Copper ion-binding proteins play an essential role in metabolic processes and are critical factors in many diseases, such as breast cancer, lung cancer, and Menkes disease. Many algorithms have been developed for predicting metal ion classification and binding sites, but none have been applied to copper ion-binding proteins. In this study, we developed a copper ion-bound protein classifier, RPCIBP, which integrating the reduced amino acid composition into position-specific score matrix (PSSM). The reduced amino acid composition filters out a large number of useless evolutionary features, improving the operational efficiency and predictive ability of the model (feature dimension from 2900 to 200, ACC from 83 % to 85.1 %). Compared with the basic model using only three sequence feature extraction methods (ACC in training set between 73.8 %-86.2 %, ACC in test set between 69.3 %-87.5 %), the model integrating the evolutionary features of the reduced amino acid composition showed higher accuracy and robustness (ACC in training set between 83.1 %-90.8 %, ACC in test set between 79.1 %-91.9 %). Best copper ion-binding protein classifiers filtered by feature selection progress were deployed in a user-friendly web server (http://bioinfor.imu.edu.cn/RPCIBP). RPCIBP can accurately predict copper ion-binding proteins, which is convenient for further structural and functional studies, and conducive to mechanism exploration and target drug development.

Strategies for activity analysis of single nucleotide polymorphisms associated with human diseases

Genome-wide association studies (GWAS) have identified a large number of single nucleotide polymorphism (SNP) sites associated with human diseases. In the annotation of human diseases, especially cancers, SNPs, as an important component of genetic factors, have gained increasing attention. Given that most of the SNPs are located in non-coding regions, the functional verification of these SNPs is a great challenge. The key to functional annotation for risk SNPs is to screen SNPs with regulatory activity from thousands of disease associated-SNPs. In this review, we systematically recapitulate the characteristics and functional roles of SNP sites, discuss three parallel reporter screening strategies in detail based on barcode tag classification, and recommend the common in silico strategies to help supplement the annotation of SNP sites with epigenetic activity analysis, prediction of target genes and trans-acting factors. We hope that this review will contribute to this exuberant research field by providing robust activity analysis strategies that can facilitate the translation of GWAS results into personalized diagnosis and prevention measures for human diseases.

Prediction of Enhancers in DNA Sequence Data using a Hybrid CNN-DLSTM Model

Enhancer, a distal cis-regulatory element controls gene expression. Experimental prediction of enhancer elements is time-consuming and expensive. Consequently, various inexpensive deep learning-based fast methods have been developed for predicting the enhancers and determining their strength. In this paper, we have proposed a two-stage deep learning-based framework leveraging DNA structural features, natural language processing, convolutional neural network, and long short-term memory to predict the enhancer elements accurately in the genomics data. In the first stage, we extracted the features from DNA sequence data by using three feature representation techniques viz., k-mer based feature extraction along with word2vector based interpretation of underlined patterns, one-hot encoding, and the DNAshape technique. In the second stage, strength of enhancers is predicted from the extracted features using a hybrid deep learning model. The method is capable of adapting itself to varying sizes of datasets. Also, as proposed model can capture long-range sequencing patterns, the robustness of the method remains unaffected against minor variations in the genomics sequence. The method outperforms the other state-of-the-art methods at both stages in terms of performance metrics of prediction accuracy, specificity, Mathew's correlation coefficient, and area under the ROC curve. In summary, the proposed method is a reliable method for enhancer prediction.

Leveraging massively parallel reporter assays for evolutionary questions

A long-standing goal of evolutionary biology is to decode how gene regulation contributes to organismal diversity. Doing so is challenging because it is hard to predict function from non-coding sequence and to perform molecular research with non-model taxa. Massively parallel reporter assays (MPRAs) enable the testing of thousands to millions of sequences for regulatory activity simultaneously. Here, we discuss the execution, advantages, and limitations of MPRAs, with a focus on evolutionary questions. We propose solutions for extending MPRAs to rare taxa and those with limited genomic resources, and we underscore MPRA's broad potential for driving genome-scale, functional studies across organisms.

Systematic assays and resources for the functional annotation of non-coding variants

Identification of genetic variation in individual genomes is now a routine procedure in human genetic research and diagnostics. For many variants, however, insufficient evidence is available to establish a pathogenic effect, particularly for variants in non-coding regions. Furthermore, the sheer number of candidate variants renders testing in individual assays virtually impossible. While scalable approaches are being developed, the selection of methods and resources and the application of a given framework to a particular disease or trait remain major challenges. This limits the translation of results from both genome-wide association studies and genome sequencing. Here, we discuss computational and experimental approaches available for functional annotation of non-coding variation.

Genome-wide identification and characterization of DNA enhancers with a stacked multivariate fusion framework

Enhancers are short non-coding DNA sequences outside of the target promoter regions that can be bound by specific proteins to increase a gene's transcriptional activity, which has a crucial role in the spatiotemporal and quantitative regulation of gene expression. However, enhancers do not have a specific sequence motifs or structures, and their scattered distribution in the genome makes the identification of enhancers from human cell lines particularly challenging. Here we present a novel, stacked multivariate fusion framework called SMFM, which enables a comprehensive identification and analysis of enhancers from regulatory DNA sequences as well as their interpretation. Specifically, to characterize the hierarchical relationships of enhancer sequences, multi-source biological information and dynamic semantic information are fused to represent regulatory DNA enhancer sequences. Then, we implement a deep learning-based sequence network to learn the feature representation of the enhancer sequences comprehensively and to extract the implicit relationships in the dynamic semantic information. Ultimately, an ensemble machine learning classifier is trained based on the refined multi-source features and dynamic implicit relations obtained from the deep learning-based sequence network. Benchmarking experiments demonstrated that SMFM significantly outperforms other existing methods using several evaluation metrics. In addition, an independent test set was used to validate the generalization performance of SMFM by comparing it to other state-of-the-art enhancer identification methods. Moreover, we performed motif analysis based on the contribution scores of different bases of enhancer sequences to the final identification results. Besides, we conducted interpretability analysis of the identified enhancer sequences based on attention weights of EnhancerBERT, a fine-tuned BERT model that provides new insights into exploring the gene semantic information likely to underlie the discovered enhancers in an interpretable manner. Finally, in a human placenta study with 4,562 active distal gene regulatory enhancers, SMFM successfully exposed tissue-related placental development and the differential mechanism, demonstrating the generalizability and stability of our proposed framework.

Multiplexed functional genomic assays to decipher the noncoding genome

Linkage disequilibrium and the incomplete regulatory annotation of the noncoding genome complicates the identification of functional noncoding genetic variants and their causal association with disease. Current computational methods for variant prioritization have limited predictive value, necessitating the application of highly parallelized experimental assays to efficiently identify functional noncoding variation. Here, we summarize two distinct approaches, massively parallel reporter assays and CRISPR-based pooled screens and describe their flexible implementation to characterize human noncoding genetic variation at unprecedented scale. Each approach provides unique advantages and limitations, highlighting the importance of multimodal methodological integration. These multiplexed assays of variant effects are undoubtedly poised to play a key role in the experimental characterization of noncoding genetic risk, informing our understanding of the underlying mechanisms of disease-associated loci and the development of more robust predictive classification algorithms.

Machine Learning Prediction of Non-Coding Variant Impact in Human Retinal cis-Regulatory Elements

Purpose

Prior studies have demonstrated the significance of specific cis-regulatory variants in retinal disease; however, determining the functional impact of regulatory variants remains a major challenge. In this study, we utilized a machine learning approach, trained on epigenomic data from the adult human retina, to systematically quantify the predicted impact of cis-regulatory variants.

Methods

We used human retinal DNA accessibility data (ATAC-seq) to determine a set of 18.9k high-confidence, putative cis-regulatory elements. Eighty percent of these elements were used to train a machine learning model utilizing a gapped k-mer support vector machine-based approach. In silico saturation mutagenesis and variant scoring was applied to predict the functional impact of all potential single nucleotide variants within cis-regulatory elements. Impact scores were tested in a 20% hold-out dataset and compared to allele population frequency, phylogenetic conservation, transcription factor (TF) binding motifs, and existing massively parallel reporter assay data.

Results

We generated a model that distinguishes between human retinal regulatory elements and negative test sequences with 95% accuracy. Among a hold-out test set of 3.7k human retinal CREs, all possible single nucleotide variants were scored. Variants with negative impact scores correlated with higher phylogenetic conservation of the reference allele, disruption of predicted TF binding motifs, and massively parallel reporter expression.

Conclusions

We demonstrated the utility of human retinal epigenomic data to train a machine learning model for the purpose of predicting the impact of non-coding regulatory sequence variants. Our model accurately scored sequences and predicted putative transcription factor binding motifs. This approach has the potential to expedite the characterization of pathogenic non-coding sequence variants in the context of unexplained retinal disease.

Translational Relevance

This workflow and resulting dataset serve as a promising genomic tool to facilitate the clinical prioritization of functionally disruptive non-coding mutations in the retina.

SemanticCAP: Chromatin Accessibility Prediction Enhanced by Features Learning from a Language Model

A large number of inorganic and organic compounds are able to bind DNA and form complexes, among which drug-related molecules are important. Chromatin accessibility changes not only directly affect drug–DNA interactions, but they can promote or inhibit the expression of the critical genes associated with drug resistance by affecting the DNA binding capacity of TFs and transcriptional regulators. However, the biological experimental techniques for measuring it are expensive and time-consuming. In recent years, several kinds of computational methods have been proposed to identify accessible regions of the genome. Existing computational models mostly ignore the contextual information provided by the bases in gene sequences. To address these issues, we proposed a new solution called SemanticCAP. It introduces a gene language model that models the context of gene sequences and is thus able to provide an effective representation of a certain site in a gene sequence. Basically, we merged the features provided by the gene language model into our chromatin accessibility model. During the process, we designed methods called SFA and SFC to make feature fusion smoother. Compared to DeepSEA, gkm-SVM, and k-mer using public benchmarks, our model proved to have better performance, showing a 1.25% maximum improvement in auROC and a 2.41% maximum improvement in auPRC.

Prioritization of regulatory variants with tissue-specific function in the non-coding regions of human genome

Understanding the functional consequences of genetic variation in the non-coding regions of the human genome remains a challenge. We introduce h ere a computational tool, TURF, to prioritize regulatory variants with tissue-specific function by leveraging evidence from functional genomics experiments, including over 3000 functional genomics datasets from the ENCODE project provided in the RegulomeDB database. TURF is able to generate prediction scores at both organism and tissue/organ-specific levels for any non-coding variant on the genome. We present that TURF has an overall top performance in prediction by using validated variants from MPRA experiments. We also demonstrate how TURF can pick out the regulatory variants with tissue-specific function over a candidate list from associate studies. Furthermore, we found that various GWAS traits showed the enrichment of regulatory variants predicted by TURF scores in the trait-relevant organs, which indicates that these variants can be a valuable source for future studies.

Linking genome variants to disease: scalable approaches to test the functional impact of human mutations

The application of genomics to medicine has accelerated the discovery of mutations underlying disease and has enhanced our knowledge of the molecular underpinnings of diverse pathologies. As the amount of human genetic material queried via sequencing has grown exponentially in recent years, so too has the number of rare variants observed. Despite progress, our ability to distinguish which rare variants have clinical significance remains limited. Over the last decade, however, powerful experimental approaches have emerged to characterize variant effects orders of magnitude faster than before. Fueled by improved DNA synthesis and sequencing and, more recently, by CRISPR/Cas9 genome editing, multiplex functional assays provide a means of generating variant effect data in wide-ranging experimental systems. Here, I review recent applications of multiplex assays that link human variants to disease phenotypes and I describe emerging strategies that will enhance their clinical utility in coming years.

Identification of Disease-Related 2-Oxoglutarate/Fe (II)-Dependent Oxygenase Based on Reduced Amino Acid Cluster Strategy

The 2-oxoglutarate/Fe (II)-dependent (2OG) oxygenase superfamily is mainly responsible for protein modification, nucleic acid repair and/or modification, and fatty acid metabolism and plays important roles in cancer, cardiovascular disease, and other diseases. They are likely to become new targets for the treatment of cancer and other diseases, so the accurate identification of 2OG oxygenases is of great significance. Many computational methods have been proposed to predict functional proteins to compensate for the time-consuming and expensive experimental identification. However, machine learning has not been applied to the study of 2OG oxygenases. In this study, we developed OGFE_RAAC, a prediction model to identify whether a protein is a 2OG oxygenase. To improve the performance of OGFE_RAAC, 673 amino acid reduction alphabets were used to determine the optimal feature representation scheme by recoding the protein sequence. The 10-fold cross-validation test showed that the accuracy of the model in identifying 2OG oxygenases is 91.04%. Besides, the independent dataset results also proved that the model has excellent generalization and robustness. It is expected to become an effective tool for the identification of 2OG oxygenases. With further research, we have also found that the function of 2OG oxygenases may be related to their polarity and hydrophobicity, which will help the follow-up study on the catalytic mechanism of 2OG oxygenases and the way they interact with the substrate. Based on the model we built, a user-friendly web server was established and can be friendly accessed at http://bioinfor.imu.edu.cn/ogferaac.

Chromatin Landscapes of Human Lung Cells Predict Potentially Functional Chronic Obstructive Pulmonary Disease Genome-Wide Association Study Variants

Genome-wide association studies (GWASs) have identified dozens of loci associated with risk of chronic obstructive pulmonary disease (COPD). However, identifying the causal variants and their functional role in the appropriate cell type remains a major challenge. We aimed to identify putative causal variants in 82 GWAS loci associated with COPD susceptibility and predict the regulatory impact of these variants in lung-cell types. We used an integrated approach featuring statistical fine mapping, open chromatin profiling, and machine learning to identify functional variants. We generated chromatin accessibility data using the Assay for Transposase-Accessible Chromatin with High-Throughput Sequencing (ATAC-seq) for human primary lung-cell types implicated in COPD pathobiology. We then evaluated the enrichment of COPD risk variants in lung-specific open chromatin regions and generated cell type-specific regulatory predictions for >6,500 variants corresponding to 82 COPD GWAS loci. Integration of the fine-mapped variants with lung open chromatin regions helped prioritize 22 variants in putative regulatory elements with potential functional effects. Comparison with functional predictions from 222 Encyclopedia of DNA Elements (ENCODE) cell samples revealed cell type-specific regulatory effects of COPD variants in the lung epithelium, endothelium, and immune cells. We identified potential causal variants for COPD risk by integrating fine mapping in GWAS loci with cell-specific regulatory profiling, highlighting the importance of leveraging the chromatin status in relevant cell types to predict the molecular effects of risk variants in lung disease.

Regulatory SNPs: Altered Transcription Factor Binding Sites Implicated in Complex Traits and Diseases

The vast majority of the genetic variants (mainly SNPs) associated with various human traits and diseases map to a noncoding part of the genome and are enriched in its regulatory compartment, suggesting that many causal variants may affect gene expression. The leading mechanism of action of these SNPs consists in the alterations in the transcription factor binding via creation or disruption of transcription factor binding sites (TFBSs) or some change in the affinity of these regulatory proteins to their cognate sites. In this review, we first focus on the history of the discovery of regulatory SNPs (rSNPs) and systematized description of the existing methodical approaches to their study. Then, we brief the recent comprehensive examples of rSNPs studied from the discovery of the changes in the TFBS sequence as a result of a nucleotide substitution to identification of its effect on the target gene expression and, eventually, to phenotype. We also describe state-of-the-art genome-wide approaches to identification of regulatory variants, including both making molecular sense of genome-wide association studies (GWAS) and the alternative approaches the primary goal of which is to determine the functionality of genetic variants. Among these approaches, special attention is paid to expression quantitative trait loci (eQTLs) analysis and the search for allele-specific events in RNA-seq (ASE events) as well as in ChIP-seq, DNase-seq, and ATAC-seq (ASB events) data.

The impact of different negative training data on regulatory sequence predictions

Regulatory regions, like promoters and enhancers, cover an estimated 5–15% of the human genome. Changes to these sequences are thought to underlie much of human phenotypic variation and a substantial proportion of genetic causes of disease. However, our understanding of their functional encoding in DNA is still very limited. Applying machine or deep learning methods can shed light on this encoding and gapped k-mer support vector machines (gkm-SVMs) or convolutional neural networks (CNNs) are commonly trained on putative regulatory sequences. Here, we investigate the impact of negative sequence selection on model performance. By training gkm-SVM and CNN models on open chromatin data and corresponding negative training dataset, both learners and two approaches for negative training data are compared. Negative sets use either genomic background sequences or sequence shuffles of the positive sequences. Model performance was evaluated on three different tasks: predicting elements active in a cell-type, predicting cell-type specific elements, and predicting elements' relative activity as measured from independent experimental data. Our results indicate strong effects of the negative training data, with genomic backgrounds showing overall best results. Specifically, models trained on highly shuffled sequences perform worse on the complex tasks of tissue-specific activity and quantitative activity prediction, and seem to learn features of artificial sequences rather than regulatory activity. Further, we observe that insufficient matching of genomic background sequences results in model biases. While CNNs achieved and exceeded the performance of gkm-SVMs for larger training datasets, gkm-SVMs gave robust and best results for typical training dataset sizes without the need of hyperparameter optimization.

Enhancer Predictions and Genome-Wide Regulatory Circuits

Spatiotemporal control of gene expression during development requires orchestrated activities of numerous enhancers, which are cis-regulatory DNA sequences that, when bound by transcription factors, support selective activation or repression of associated genes. Proper activation of enhancers is critical during embryonic development, adult tissue homeostasis, and regeneration, and inappropriate enhancer activity is often associated with pathological conditions such as cancer. Multiple consortia [e.g., the Encyclopedia of DNA Elements (ENCODE) Consortium and National Institutes of Health Roadmap Epigenomics Mapping Consortium] and independent investigators have mapped putative regulatory regions in a large number of cell types and tissues, but the sequence determinants of cell-specific enhancers are not yet fully understood. Machine learning approaches trained on large sets of these regulatory regions can identify core transcription factor binding sites and generate quantitative predictions of enhancer activity and the impact of sequence variants on activity. Here, we review these computational methods in the context of enhancer prediction and gene regulatory network models specifying cell fate.

A Pretraining-Retraining Strategy of Deep Learning Improves Cell-Specific Enhancer Predictions

Deciphering the code of cis-regulatory element (CRE) is one of the core issues of today’s biology. Enhancers are distal CREs and play significant roles in gene transcriptional regulation. Although identifications of enhancer locations across the whole genome [discriminative enhancer predictions (DEP)] is necessary, it is more important to predict in which specific cell or tissue types, they will be activated and functional [tissue-specific enhancer predictions (TSEP)]. Although existing deep learning models achieved great successes in DEP, they cannot be directly employed in TSEP because a specific cell or tissue type only has a limited number of available enhancer samples for training. Here, we first adopted a reported deep learning architecture and then developed a novel training strategy named “pretraining-retraining strategy” (PRS) for TSEP by decomposing the whole training process into two successive stages: a pretraining stage is designed to train with the whole enhancer data for performing DEP, and a retraining strategy is then designed to train with tissue-specific enhancer samples based on the trained pretraining model for making TSEP. As a result, PRS is found to be valid for DEP with an AUC of 0.922 and a GM (geometric mean) of 0.696, when testing on a larger-scale FANTOM5 enhancer dataset via a five-fold cross-validation. Interestingly, based on the trained pretraining model, a new finding is that only additional twenty epochs are needed to complete the retraining process on testing 23 specific tissues or cell lines. For TSEP tasks, PRS achieved a mean GM of 0.806 which is significantly higher than 0.528 of gkm-SVM, an existing mainstream method for CRE predictions. Notably, PRS is further proven superior to other two state-of-the-art methods: DEEP and BiRen. In summary, PRS has employed useful ideas from the domain of transfer learning and is a reliable method for TSEPs.

Meta-analysis of massively parallel reporter assays enables prediction of regulatory function across cell types

Deciphering the potential of noncoding loci to influence gene regulation has been the subject of intense research, with important implications in understanding genetic underpinnings of human diseases. Massively parallel reporter assays (MPRAs) can measure regulatory activity of thousands of DNA sequences and their variants in a single experiment. With increasing number of publically available MPRA data sets, one can now develop data-driven models which, given a DNA sequence, predict its regulatory activity. Here, we performed a comprehensive meta-analysis of several MPRA data sets in a variety of cellular contexts. We first applied an ensemble of methods to predict MPRA output in each context and observed that the most predictive features are consistent across data sets. We then demonstrate that predictive models trained in one cellular context can be used to predict MPRA output in another, with loss of accuracy attributed to cell-type-specific features. Finally, we show that our approach achieves top performance in the Fifth Critical Assessment of Genome Interpretation "Regulation Saturation" Challenge for predicting effects of single-nucleotide variants. Overall, our analysis provides insights into how MPRA data can be leveraged to highlight functional regulatory regions throughout the genome and can guide effective design of future experiments by better prioritizing regions of interest.

Integration of multiple epigenomic marks improves prediction of variant impact in saturation mutagenesis reporter assay

The integrative analysis of high-throughput reporter assays, machine learning, and profiles of epigenomic chromatin state in a broad array of cells and tissues has the potential to significantly improve our understanding of noncoding regulatory element function and its contribution to human disease. Here, we report results from the CAGI 5 regulation saturation challenge where participants were asked to predict the impact of nucleotide substitution at every base pair within five disease-associated human enhancers and nine disease-associated promoters. A library of mutations covering all bases was generated by saturation mutagenesis and altered activity was assessed in a massively parallel reporter assay (MPRA) in relevant cell lines. Reporter expression was measured relative to plasmid DNA to determine the impact of variants. The challenge was to predict the functional effects of variants on reporter expression. Comparative analysis of the full range of submitted prediction results identifies the most successful models of transcription factor binding sites, machine learning algorithms, and ways to choose among or incorporate diverse datatypes and cell-types for training computational models. These results have the potential to improve the design of future studies on more diverse sets of regulatory elements and aid the interpretation of disease-associated genetic variation.

TAGOOS: genome-wide supervised learning of non-coding loci associated to complex phenotypes

Genome-wide association studies (GWAS) associate single nucleotide polymorphisms (SNPs) to complex phenotypes. Most human SNPs fall in non-coding regions and are likely regulatory SNPs, but linkage disequilibrium (LD) blocks make it difficult to distinguish functional SNPs. Therefore, putative functional SNPs are usually annotated with molecular markers of gene regulatory regions and prioritized with dedicated prediction tools. We integrated associated SNPs, LD blocks and regulatory features into a supervised model called TAGOOS (TAG SNP bOOSting) and computed scores genome-wide. The TAGOOS scores enriched and prioritized unseen associated SNPs with an odds ratio of 4.3 and 3.5 and an area under the curve (AUC) of 0.65 and 0.6 for intronic and intergenic regions, respectively. The TAGOOS score was correlated with the maximal significance of associated SNPs and expression quantitative trait loci (eQTLs) and with the number of biological samples annotated for key regulatory features. Analysis of loci and regions associated to cleft lip and human adult height phenotypes recovered known functional loci and predicted new functional loci enriched in transcriptions factors related to the phenotypes. In conclusion, we trained a supervised model based on associated SNPs to prioritize putative functional regions. The TAGOOS scores, annotations and UCSC genome tracks are available here: https://tagoos.readthedocs.io.

sefOri: selecting the best-engineered sequence features to predict DNA replication origins

Motivation

Cell divisions start from replicating the double-stranded DNA, and the DNA replication process needs to be precisely regulated both spatially and temporally. The DNA is replicated starting from the DNA replication origins. A few successful prediction models were generated based on the assumption that the DNA replication origin regions have sequence level features like physicochemical properties significantly different from the other DNA regions.

Results

This study proposed a feature selection procedure to further refine the classification model of the DNA replication origins. The experimental data demonstrated that as large as 26% improvement in the prediction accuracy may be achieved on the yeast Saccharomyces cerevisiae. Moreover, the prediction accuracies of the DNA replication origins were improved for all the four yeast genomes investigated in this study.

Availability and implementation

The software sefOri version 1.0 was available at http://www.healthinformaticslab.org/supp/resources.php. An online server was also provided for the convenience of the users, and its web link may be found in the above-mentioned web page.

Supplementary information

Supplementary data are available at Bioinformatics online.

Deciphering regulatory DNA sequences and noncoding genetic variants using neural network models of massively parallel reporter assays

The relationship between noncoding DNA sequence and gene expression is not well-understood. Massively parallel reporter assays (MPRAs), which quantify the regulatory activity of large libraries of DNA sequences in parallel, are a powerful approach to characterize this relationship. We present MPRA-DragoNN, a convolutional neural network (CNN)-based framework to predict and interpret the regulatory activity of DNA sequences as measured by MPRAs. While our method is generally applicable to a variety of MPRA designs, here we trained our model on the Sharpr-MPRA dataset that measures the activity of ∼500,000 constructs tiling 15,720 regulatory regions in human K562 and HepG2 cell lines. MPRA-DragoNN predictions were moderately correlated (Spearman ρ = 0.28) with measured activity and were within range of replicate concordance of the assay. State-of-the-art model interpretation methods revealed high-resolution predictive regulatory sequence features that overlapped transcription factor (TF) binding motifs. We used the model to investigate the cell type and chromatin state preferences of predictive TF motifs. We explored the ability of our model to predict the allelic effects of regulatory variants in an independent MPRA experiment and fine map putative functional SNPs in loci associated with lipid traits. Our results suggest that interpretable deep learning models trained on MPRA data have the potential to reveal meaningful patterns in regulatory DNA sequences and prioritize regulatory genetic variants, especially as larger, higher-quality datasets are produced.

Genome-scale screens identify JNK-JUN signaling as a barrier for pluripotency exit and endoderm differentiation

Human embryonic and induced pluripotent stem cells (hESCs/hiPSCs) hold great promise for cell-based therapies and drug discovery. However, homogeneous differentiation remains a major challenge, highlighting the need for understanding developmental mechanisms. We performed genome-scale CRISPR screens to uncover regulators of definitive endoderm (DE) differentiation, which unexpectedly uncovered five JNK/JUN family genes as key barriers of DE differentiation. The JNK/JUN pathway does not act through directly inhibiting the DE enhancers. Instead JUN co-occupies ESC enhancers with OCT4, NANOG and SMAD2/3, and specifically inhibits the exit from the pluripotent state by impeding the decommissioning of ESC enhancers and inhibiting the reconfiguration of SMAD2/3 chromatin binding from ESC to DE enhancers. Therefore, the JNK/JUN pathway safeguards pluripotency from precocious DE differentiation. Direct pharmacological inhibition of JNK significantly improves the efficiencies of generating DE and DE-derived pancreatic and lung progenitor cells, highlighting the potential of harnessing the knowledge from developmental studies for regenerative medicine.

Improved Prediction of Regulatory Element Using Hybrid Abelian Complexity Features with DNA Sequences

Deciphering the code of cis-regulatory element (CRE) is one of the core issues of current biology. As an important category of CRE, enhancers play crucial roles in gene transcriptional regulations in a distant manner. Further, the disruption of an enhancer can cause abnormal transcription and, thus, trigger human diseases, which means that its accurate identification is currently of broad interest. Here, we introduce an innovative concept, i.e., abelian complexity function (ACF), which is a more complex extension of the classic subword complexity function, for a new coding of DNA sequences. After feature selection by an upper bound estimation and integration with DNA composition features, we developed an enhancer prediction model with hybrid abelian complexity features (HACF). Compared with existing methods, HACF shows consistently superior performance on three sources of enhancer datasets. We tested the generalization ability of HACF by scanning human chromosome 22 to validate previously reported super-enhancers. Meanwhile, we identified novel candidate enhancers which have supports from enhancer-related ENCODE ChIP-seq signals. In summary, HACF improves current enhancer prediction and may be beneficial for further prioritization of functional noncoding variants.

Human cardiac cis-regulatory elements, their cognate transcription factors, and regulatory DNA sequence variants

Cis-regulatory elements (CRE), short DNA sequences through which transcription factors (TFs) exert regulatory control on gene expression, are postulated to be the major sites of causal sequence variation underlying the genetics of complex traits and diseases. We present integrative analyses, combining high-throughput genomic and epigenomic data with sequence-based computations, to identify the causal transcriptional components in a given tissue. We use data on adult human hearts to demonstrate that (1) sequence-based predictions detect numerous, active, tissue-specific CREs missed by experimental observations, (2) learned sequence features identify the cognate TFs, (3) CRE variants are specifically associated with cardiac gene expression, and (4) a significant fraction of the heritability of exemplar cardiac traits (QT interval, blood pressure, pulse rate) is attributable to these variants. This general systems approach can thus identify candidate causal variants and the components of gene regulatory networks (GRN) to enable understanding of the mechanisms of complex disorders on a tissue- or cell-type basis.

Functional characterization of enhancer evolution in the primate lineage

BACKGROUND

Enhancers play an important role in morphological evolution and speciation by controlling the spatiotemporal expression of genes. Previous efforts to understand the evolution of enhancers in primates have typically studied many enhancers at low resolution, or single enhancers at high resolution. Although comparative genomic studies reveal large-scale turnover of enhancers, a specific understanding of the molecular steps by which mammalian or primate enhancers evolve remains elusive.

RESULTS

We identified candidate hominoid-specific liver enhancers from H3K27ac ChIP-seq data. After locating orthologs in 11 primates spanning around 40 million years, we synthesized all orthologs as well as computational reconstructions of 9 ancestral sequences for 348 active tiles of 233 putative enhancers. We concurrently tested all sequences for regulatory activity with STARR-seq in HepG2 cells. We observe groups of enhancer tiles with coherent trajectories, most of which can be potentially explained by a single gain or loss-of-activity event per tile. We quantify the correlation between the number of mutations along a branch and the magnitude of change in functional activity. Finally, we identify 84 mutations that correlate with functional changes; these are enriched for cytosine deamination events within CpGs.

CONCLUSIONS

We characterized the evolutionary-functional trajectories of hundreds of liver enhancers throughout the primate phylogeny. We observe subsets of regulatory sequences that appear to have gained or lost activity. We use these data to quantify the relationship between sequence and functional divergence, and to identify CpG deamination as a potentially important force in driving changes in enhancer activity during primate evolution.

Limb-Enhancer Genie: An accessible resource of accurate enhancer predictions in the developing limb

Epigenomic mapping of enhancer-associated chromatin modifications facilitates the genome-wide discovery of tissue-specific enhancers in vivo. However, reliance on single chromatin marks leads to high rates of false-positive predictions. More sophisticated, integrative methods have been described, but commonly suffer from limited accessibility to the resulting predictions and reduced biological interpretability. Here we present the Limb-Enhancer Genie (LEG), a collection of highly accurate, genome-wide predictions of enhancers in the developing limb, available through a user-friendly online interface. We predict limb enhancers using a combination of >50 published limb-specific datasets and clusters of evolutionarily conserved transcription factor binding sites, taking advantage of the patterns observed at previously in vivo validated elements. By combining different statistical models, our approach outperforms current state-of-the-art methods and provides interpretable measures of feature importance. Our results indicate that including a previously unappreciated score that quantifies tissue-specific nuclease accessibility significantly improves prediction performance. We demonstrate the utility of our approach through in vivo validation of newly predicted elements. Moreover, we describe general features that can guide the type of datasets to include when predicting tissue-specific enhancers genome-wide, while providing an accessible resource to the general biological community and facilitating the functional interpretation of genetic studies of limb malformations.