Deep learning of cross-species single cell atlases identifies conserved regulatory programs underlying cell types

Despite extensive efforts to sequence different genomes, genetic models to interpret gene regulation and cell fate decisions are lacking for most species. Here, we performed whole-body single-cell transcriptomic analysis of zebrafish, Drosophila, and earthworm. We then mapped cell landscapes covering eight representative metazoan species to study gene regulation through evolution. With uniformly constructed cross-species datasets, we developed a deep learning-based strategy, Nvwa, to predict gene expression landscapes and decipher regulatory sequences at the single-cell level. We systematically compared cell type-specific transcription factors (TFs) to reveal conserved genetic regulation among vertebrates and invertebrates. Our work provides a valuable resource and a novel strategy for studying regulatory language in diverse biological systems.