De novo distillation of thermodynamic affinity from deep learning regulatory sequence models of in vivo protein-DNA binding.

Transcription factors (TF) are proteins that bind DNA in a sequence-specific manner to regulate gene transcription. Despite their unique intrinsic sequence preferences, in vivo genomic occupancy profiles of TFs differ across cellular contexts. Hence, deciphering the sequence determinants of TF binding, both intrinsic and context-specific, is essential to understand gene regulation and the impact of regulatory, non-coding genetic variation. Biophysical models trained on in vitro TF binding assays can estimate intrinsic affinity landscapes and predict occupancy based on TF concentration and affinity. However, these models cannot adequately explain context-specific, in vivo binding profiles. Conversely, deep learning models, trained on in vivo TF binding assays, effectively predict and explain genomic occupancy profiles as a function of complex regulatory sequence syntax, albeit without a clear biophysical interpretation. To reconcile these complementary models of in vitro and in vivo TF binding, we developed Affinity Distillation (AD), a method that extracts thermodynamic affinities de-novo from deep learning models of TF chromatin immunoprecipitation (ChIP) experiments by marginalizing away the influence of genomic sequence context. Applied to neural networks modeling diverse classes of yeast and mammalian TFs, AD predicts energetic impacts of sequence variation within and surrounding motifs on TF binding as measured by diverse in vitro assays with superior dynamic range and accuracy compared to motif-based methods. Furthermore, AD can accurately discern affinities of TF paralogs. Our results highlight thermodynamic affinity as a key determinant of in vivo binding, suggest that deep learning models of in vivo binding implicitly learn high-resolution affinity landscapes, and show that these affinities can be successfully distilled using AD. This new biophysical interpretation of deep learning models enables high-throughput in silico experiments to explore the influence of sequence context and variation on both intrinsic affinity and in vivo occupancy.

Chromatin accessibility in the Drosophila embryo is determined by transcription factor pioneering and enhancer activation.

Chromatin accessibility is integral to the process by which transcription factors (TFs) read out cis-regulatory DNA sequences, but it is difficult to differentiate between TFs that drive accessibility and those that do not. Deep learning models that learn complex sequence rules provide an unprecedented opportunity to dissect this problem. Using zygotic genome activation in Drosophila as a model, we analyzed high-resolution TF binding and chromatin accessibility data with interpretable deep learning and performed genetic validation experiments. We identify a hierarchical relationship between the pioneer TF Zelda and the TFs involved in axis patterning. Zelda consistently pioneers chromatin accessibility proportional to motif affinity, whereas patterning TFs augment chromatin accessibility in sequence contexts where they mediate enhancer activation. We conclude that chromatin accessibility occurs in two tiers: one through pioneering, which makes enhancers accessible but not necessarily active, and the second when the correct combination of TFs leads to enhancer activation.

The multi-lineage transcription factor ISL1 controls cardiomyocyte cell fate through interaction with NKX2.5.

Congenital heart disease often arises from perturbations of transcription factors (TFs) that guide cardiac development. ISLET1 (ISL1) is a TF that influences early cardiac cell fate, as well as differentiation of other cell types including motor neuron progenitors (MNPs) and pancreatic islet cells. While lineage specificity of ISL1 function is likely achieved through combinatorial interactions, its essential cardiac interacting partners are unknown. By assaying ISL1 genomic occupancy in human induced pluripotent stem cell-derived cardiac progenitors (CPs) or MNPs and leveraging the deep learning approach BPNet, we identified motifs of other TFs that predicted ISL1 occupancy in each lineage, with NKX2.5 and GATA motifs being most closely associated to ISL1 in CPs. Experimentally, nearly two-thirds of ISL1-bound loci were co-occupied by NKX2.5 and/or GATA4. Removal of NKX2.5 from CPs led to widespread ISL1 redistribution, and overexpression of NKX2.5 in MNPs led to ISL1 occupancy of CP-specific loci. These results reveal how ISL1 guides lineage choices through a combinatorial code that dictates genomic occupancy and transcription.

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

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