MUSE enables cross-species multi-omics integration that incorporates transcriptional regulatory modules

Recent advances in evolutionary biology and biomedical research have promoted comparative analyses of cellular states and developmental processes across species, leading to the development of numerous cross-species alignment methods based on scRNA-seq data. However, alignments relying solely on RNA expression are strongly driven by lineage signals and cell typespecific transcriptional programs. As a result, they are limited in their ability to identify conserved regulatory modules across species and to compare regulatory logic beyond developmental lineages, thereby constraining biological interpretability. Meanwhile, recent technological developments have enabled the acquisition of multi-omics data, including chromatin accessibility, making it increasingly feasible to analyze and interpret conservation at the level of regulatory modules across species. Nevertheless, computational methods that can integratively handle such heterogeneous omics data and enable cross-species comparative analysis in a unified framework remain insufficiently established. To address this challenge, we propose Multi-omics Unified embedding across Species (MUSE), a novel framework for integrating multi-omics data across species. MUSE constructs a graph that captures relationships among features both within and across species, and learns a shared latent space based on this graph structure. By leveraging this integrated graph-based representation, MUSE enables cross-species alignment that preserves species-specific characteristics while reflecting similarities not only at the level of gene expression and chromatin states but also at the level of regulatory modules.