Prediction of gene regulatory enhancers across species reveals evolutionarily conserved sequence properties

Cross-species analysis of enhancer logic using deep learning

Deciphering the genomic regulatory code of enhancers is a key challenge in biology because this code underlies cellular identity. A better understanding of how enhancers work will improve the interpretation of noncoding genome variation and empower the generation of cell type-specific drivers for gene therapy. Here, we explore the combination of deep learning and cross-species chromatin accessibility profiling to build explainable enhancer models. We apply this strategy to decipher the enhancer code in melanoma, a relevant case study owing to the presence of distinct melanoma cell states. We trained and validated a deep learning model, called DeepMEL, using chromatin accessibility data of 26 melanoma samples across six different species. We show the accuracy of DeepMEL predictions on the CAGI5 challenge, where it significantly outperforms existing models on the melanoma enhancer of IRF4 Next, we exploit DeepMEL to analyze enhancer architectures and identify accurate transcription factor binding sites for the core regulatory complexes in the two different melanoma states, with distinct roles for each transcription factor, in terms of nucleosome displacement or enhancer activation. Finally, DeepMEL identifies orthologous enhancers across distantly related species, where sequence alignment fails, and the model highlights specific nucleotide substitutions that underlie enhancer turnover. DeepMEL can be used from the Kipoi database to predict and optimize candidate enhancers and to prioritize enhancer mutations. In addition, our computational strategy can be applied to other cancer or normal cell types.

Learning and interpreting the gene regulatory grammar in a deep learning framework

Deep neural networks (DNNs) have achieved state-of-the-art performance in identifying gene regulatory sequences, but they have provided limited insight into the biology of regulatory elements due to the difficulty of interpreting the complex features they learn. Several models of how combinatorial binding of transcription factors, i.e. the regulatory grammar, drives enhancer activity have been proposed, ranging from the flexible TF billboard model to the stringent enhanceosome model. However, there is limited knowledge of the prevalence of these (or other) sequence architectures across enhancers. Here we perform several hypothesis-driven analyses to explore the ability of DNNs to learn the regulatory grammar of enhancers. We created synthetic datasets based on existing hypotheses about combinatorial transcription factor binding site (TFBS) patterns, including homotypic clusters, heterotypic clusters, and enhanceosomes, from real TF binding motifs from diverse TF families. We then trained deep residual neural networks (ResNets) to model the sequences under a range of scenarios that reflect real-world multi-label regulatory sequence prediction tasks. We developed a gradient-based unsupervised clustering method to extract the patterns learned by the ResNet models. We demonstrated that simulated regulatory grammars are best learned in the penultimate layer of the ResNets, and the proposed method can accurately retrieve the regulatory grammar even when there is heterogeneity in the enhancer categories and a large fraction of TFBS outside of the regulatory grammar. However, we also identify common scenarios where ResNets fail to learn simulated regulatory grammars. Finally, we applied the proposed method to mouse developmental enhancers and were able to identify the components of a known heterotypic TF cluster. Our results provide a framework for interpreting the regulatory rules learned by ResNets, and they demonstrate that the ability and efficiency of ResNets in learning the regulatory grammar depends on the nature of the prediction task.

Cross-species regulatory sequence activity prediction

Machine learning algorithms trained to predict the regulatory activity of nucleic acid sequences have revealed principles of gene regulation and guided genetic variation analysis. While the human genome has been extensively annotated and studied, model organisms have been less explored. Model organism genomes offer both additional training sequences and unique annotations describing tissue and cell states unavailable in humans. Here, we develop a strategy to train deep convolutional neural networks simultaneously on multiple genomes and apply it to learn sequence predictors for large compendia of human and mouse data. Training on both genomes improves gene expression prediction accuracy on held out and variant sequences. We further demonstrate a novel and powerful approach to apply mouse regulatory models to analyze human genetic variants associated with molecular phenotypes and disease. Together these techniques unleash thousands of non-human epigenetic and transcriptional profiles toward more effective investigation of how gene regulation affects human disease.

Plant pangenomics: approaches, applications and advancements

With the assembly of increasing numbers of plant genomes, it is becoming accepted that a single reference assembly does not reflect the gene diversity of a species. The production of pangenomes, which reflect the structural variation and polymorphisms in genomes, enables in depth comparisons of variation within species or higher taxonomic groups. In this review, we discuss the current and emerging approaches for pangenome assembly, analysis and visualisation. In addition, we consider the potential of pangenomes for applied crop improvement, evolutionary and biodiversity studies. To fully exploit the value of pangenomes it is important to integrate broad information such as phenotypic, environmental, and expression data to gain insights into the role of variable regions within genomes.

Analysis of zebrafish periderm enhancers facilitates identification of a regulatory variant near human KRT8/18

Genome-wide association studies for non-syndromic orofacial clefting (OFC) have identified single nucleotide polymorphisms (SNPs) at loci where the presumed risk-relevant gene is expressed in oral periderm. The functional subsets of such SNPs are difficult to predict because the sequence underpinnings of periderm enhancers are unknown. We applied ATAC-seq to models of human palate periderm, including zebrafish periderm, mouse embryonic palate epithelia, and a human oral epithelium cell line, and to complementary mesenchymal cell types. We identified sets of enhancers specific to the epithelial cells and trained gapped-kmer support-vector-machine classifiers on these sets. We used the classifiers to predict the effects of 14 OFC-associated SNPs at 12q13 near KRT18. All the classifiers picked the same SNP as having the strongest effect, but the significance was highest with the classifier trained on zebrafish periderm. Reporter and deletion analyses support this SNP as lying within a periderm enhancer regulating KRT18/KRT8 expression.

How to study enhancers in non-traditional insect models

Transcriptional enhancers are central to the function and evolution of genes and gene regulation. At the organismal level, enhancers play a crucial role in coordinating tissue- and context-dependent gene expression. At the population level, changes in enhancers are thought to be a major driving force that facilitates evolution of diverse traits. An amazing array of diverse traits seen in insect morphology, physiology and behavior has been the subject of research for centuries. Although enhancer studies in insects outside of Drosophila have been limited, recent advances in functional genomic approaches have begun to make such studies possible in an increasing selection of insect species. Here, instead of comprehensively reviewing currently available technologies for enhancer studies in established model organisms such as Drosophila, we focus on a subset of computational and experimental approaches that are likely applicable to non-Drosophila insects, and discuss the pros and cons of each approach. We discuss the importance of validating enhancer function and evaluate several possible validation methods, such as reporter assays and genome editing. Key points and potential pitfalls when establishing a reporter assay system in non-traditional insect models are also discussed. We close with a discussion of how to advance enhancer studies in insects, both by improving computational approaches and by expanding the genetic toolbox in various insects. Through these discussions, this Review provides a conceptual framework for studying the function and evolution of enhancers in non-traditional insect models.

Representation learning of genomic sequence motifs with convolutional neural networks

Although convolutional neural networks (CNNs) have been applied to a variety of computational genomics problems, there remains a large gap in our understanding of how they build representations of regulatory genomic sequences. Here we perform systematic experiments on synthetic sequences to reveal how CNN architecture, specifically convolutional filter size and max-pooling, influences the extent that sequence motif representations are learned by first layer filters. We find that CNNs designed to foster hierarchical representation learning of sequence motifs—assembling partial features into whole features in deeper layers—tend to learn distributed representations, i.e. partial motifs. On the other hand, CNNs that are designed to limit the ability to hierarchically build sequence motif representations in deeper layers tend to learn more interpretable localist representations, i.e. whole motifs. We then validate that this representation learning principle established from synthetic sequences generalizes to in vivo sequences.

Although deep convolutional neural networks (CNNs) have demonstrated promise across many regulatory genomics prediction tasks, their inner workings largely remain a mystery. Here we empirically demonstrate how CNN architecture influences the extent that representations of sequence motifs are captured by first layer filters. We find that max-pooling and convolutional filter size modulates information flow, controlling the extent that deeper layers can build features hierarchically. CNNs designed to foster hierarchical representation learning tend to capture partial representations of motifs in first layer filters. On the other hand, CNNs that are designed to limit the ability of deeper layers to hierarchically build upon low-level features tend to learn whole representations of motifs in first layer filters. Together, this study enables the design of CNNs that intentionally learn interpretable representations in easier to access first layer filters (with a small tradeoff in performance), versus building harder to interpret distributed representations, both of which have their strengths and limitations.

Inherency of Form and Function in Animal Development and Evolution

I discuss recent work on the origins of morphology and cell-type diversification in Metazoa – collectively the animals – and propose a scenario for how these two properties became integrated, with the help of a third set of processes, cellular pattern formation, into the developmental programs seen in present-day metazoans. Inherent propensities to generate familiar forms and cell types, in essence a parts kit for the animals, are exhibited by present-day organisms and were likely more prominent in primitive ones. The structural motifs of animal bodies and organs, e.g., multilayered, hollow, elongated and segmented tissues, internal and external appendages, branched tubes, and modular endoskeletons, can be accounted for by the properties of mesoscale masses of metazoan cells. These material properties, in turn, resulted from the recruitment of “generic” physical forces and mechanisms – adhesion, contraction, polarity, chemical oscillation, diffusion – by toolkit molecules that were partly conserved from unicellular holozoan antecedents and partly novel, distributed in the different metazoan phyla in a fashion correlated with morphological complexity. The specialized functions of the terminally differentiated cell types in animals, e.g., contraction, excitability, barrier function, detoxification, excretion, were already present in ancestral unicellular organisms. These functions were implemented in metazoan differentiation in some cases using the same transcription factors as in single-celled ancestors, although controlled by regulatory mechanisms that were hybrids between earlier-evolved processes and regulatory innovations, such as enhancers. Cellular pattern formation, mediated by released morphogens interacting with biochemically responsive and excitable tissues, drew on inherent self-organizing processes in proto-metazoans to transform clusters of holozoan cells into animal embryos and organs.