Integrative single-cell analysis of cardiogenesis identifies developmental trajectories and non-coding mutations in congenital heart disease.

To define the multi-cellular epigenomic and transcriptional landscape of cardiac cellular development, we generated single-cell chromatin accessibility maps of human fetal heart tissues. We identified eight major differentiation trajectories involving primary cardiac cell types, each associated with dynamic transcription factor (TF) activity signatures. We contrasted regulatory landscapes of iPSC-derived cardiac cell types and their in vivo counterparts, which enabled optimization of in vitro differentiation of epicardial cells. Further, we interpreted sequence based deep learning models of cell-type-resolved chromatin accessibility profiles to decipher underlying TF motif lexicons. De novo mutations predicted to affect chromatin accessibility in arterial endothelium were enriched in congenital heart disease (CHD) cases vs. controls. In vitro studies in iPSCs validated the functional impact of identified variation on the predicted developmental cell types. This work thus defines the cell-type-resolved cis-regulatory sequence determinants of heart development and identifies disruption of cell type-specific regulatory elements in CHD.

CodonBert large language model for mRNA vaccines.

mRNA-based vaccines and therapeutics are gaining popularity and usage across a wide range of conditions. One of the critical issues when designing such mRNAs is sequence optimization. Even small proteins or peptides can be encoded by an enormously large number of mRNAs. The actual mRNA sequence can have a large impact on several properties including expression, stability, immunogenicity, and more. To enable the selection of an optimal sequence, we developed CodonBERT, a large language model (LLM) for mRNAs. Unlike prior models, CodonBERT uses codons as inputs which enables it to learn better representations. CodonBERT was trained using more than 10 million mRNA sequences from a diverse set of organisms. The resulting model captures important biological concepts. CodonBERT can also be extended to perform prediction tasks for various mRNA properties. CodonBERT outperforms previous mRNA prediction methods including on a new flu vaccine dataset.

Domain-adaptive neural networks improve cross-species prediction of transcription factor binding.

The intrinsic DNA sequence preferences and cell type-specific cooperative partners of transcription factors (TFs) are typically highly conserved. Hence, despite the rapid evolutionary turnover of individual TF binding sites, predictive sequence models of cell type-specific genomic occupancy of a TF in one species should generalize to closely matched cell types in a related species. To assess the viability of cross-species TF binding prediction, we train neural networks to discriminate ChIP-seq peak locations from genomic background and evaluate their performance within and across species. Cross-species predictive performance is consistently worse than within-species performance, which we show is caused in part by species-specific repeats. To account for this domain shift, we use an augmented network architecture to automatically discourage learning of training species-specific sequence features. This domain adaptation approach corrects for prediction errors on species-specific repeats and improves overall cross-species model performance. Our results show that cross-species TF binding prediction is feasible when models account for domain shifts driven by species-specific repeats.

Representations of lipid nanoparticles using large language models for transfection efficiency prediction.

MOTIVATION: Lipid nanoparticles (LNPs) are the most widely used vehicles for mRNA vaccine delivery. The structure of the lipids composing the LNPs can have a major impact on the effectiveness of the mRNA payload. Several properties should be optimized to improve delivery and expression including biodegradability, synthetic accessibility, and transfection efficiency. RESULTS: To optimize LNPs, we developed and tested models that enable the virtual screening of LNPs with high transfection efficiency. Our best method uses the lipid Simplified Molecular-Input Line-Entry System (SMILES) as inputs to a large language model. Large language model-generated embeddings are then used by a downstream gradient-boosting classifier. As we show, our method can more accurately predict lipid properties, which could lead to higher efficiency and reduced experimental time and costs. AVAILABILITY AND IMPLEMENTATION: Code and data links available at: https://github.com/Sanofi-Public/LipoBART.

Accelerating in silico saturation mutagenesis using compressed sensing.

MOTIVATION: In silico saturation mutagenesis (ISM) is a popular approach in computational genomics for calculating feature attributions on biological sequences that proceeds by systematically perturbing each position in a sequence and recording the difference in model output. However, this method can be slow because systematically perturbing each position requires performing a number of forward passes proportional to the length of the sequence being examined. RESULTS: In this work, we propose a modification of ISM that leverages the principles of compressed sensing to require only a constant number of forward passes, regardless of sequence length, when applied to models that contain operations with a limited receptive field, such as convolutions. Our method, named Yuzu, can reduce the time that ISM spends in convolution operations by several orders of magnitude and, consequently, Yuzu can speed up ISM on several commonly used architectures in genomics by over an order of magnitude. Notably, we found that Yuzu provides speedups that increase with the complexity of the convolution operation and the length of the sequence being analyzed, suggesting that Yuzu provides large benefits in realistic settings. AVAILABILITY AND IMPLEMENTATION: We have made this tool available at https://github.com/kundajelab/yuzu.

Transcription factor stoichiometry, motif affinity and syntax regulate single-cell chromatin dynamics during fibroblast reprogramming to pluripotency.

Ectopic expression of OCT4, SOX2, KLF4 and MYC (OSKM) transforms differentiated cells into induced pluripotent stem cells. To refine our mechanistic understanding of reprogramming, especially during the early stages, we profiled chromatin accessibility and gene expression at single-cell resolution across a densely sampled time course of human fibroblast reprogramming. Using neural networks that map DNA sequence to ATAC-seq profiles at base-resolution, we annotated cell-state-specific predictive transcription factor (TF) motif syntax in regulatory elements, inferred affinity- and concentration-dependent dynamics of Tn5-bias corrected TF footprints, linked peaks to putative target genes, and elucidated rewiring of TF-to-gene cis-regulatory networks. Our models reveal that early in reprogramming, OSK, at supraphysiological concentrations, rapidly open transient regulatory elements by occupying non-canonical low-affinity binding sites. As OSK concentration falls, the accessibility of these transient elements decays as a function of motif affinity. We find that these OSK-dependent transient elements sequester the somatic TF AP-1. This redistribution is strongly associated with the silencing of fibroblast-specific genes within individual nuclei. Together, our integrated single-cell resource and models reveal insights into the cis-regulatory code of reprogramming at unprecedented resolution, connect TF stoichiometry and motif syntax to diversification of cell fate trajectories, and provide new perspectives on the dynamics and role of transient regulatory elements in somatic silencing.

A genome-wide atlas of co-essential modules assigns function to uncharacterized genes.

A central question in the post-genomic era is how genes interact to form biological pathways. Measurements of gene dependency across hundreds of cell lines have been used to cluster genes into 'co-essential' pathways, but this approach has been limited by ubiquitous false positives. In the present study, we develop a statistical method that enables robust identification of gene co-essentiality and yields a genome-wide set of functional modules. This atlas recapitulates diverse pathways and protein complexes, and predicts the functions of 108 uncharacterized genes. Validating top predictions, we show that TMEM189 encodes plasmanylethanolamine desaturase, a key enzyme for plasmalogen synthesis. We also show that C15orf57 encodes a protein that binds the AP2 complex, localizes to clathrin-coated pits and enables efficient transferrin uptake. Finally, we provide an interactive webtool for the community to explore our results, which establish co-essentiality profiling as a powerful resource for biological pathway identification and discovery of new gene functions.