Deciphering universal gene regulatory mechanisms in diverse organisms holds great potential for advancing our knowledge of fundamental life processes and facilitating clinical applications. However, the traditional research paradigm primarily focuses on individual model organisms and does not integrate various cell types across species. Recent breakthroughs in single-cell sequencing and deep learning techniques present an unprecedented opportunity to address this challenge. In this study, we built an extensive dataset of over 120 million human and mouse single-cell transcriptomes. After data preprocessing, we obtained 101,768,420 single-cell transcriptomes and developed a knowledge-informed cross-species foundation model, named GeneCompass. During pre-training, GeneCompass effectively integrated four types of prior biological knowledge to enhance our understanding of gene regulatory mechanisms in a self-supervised manner. By fine-tuning for multiple downstream tasks, GeneCompass outperformed state-of-the-art models in diverse applications for a single species and unlocked new realms of cross-species biological investigations. We also employed GeneCompass to search for key factors associated with cell fate transition and showed that the predicted candidate genes could successfully induce the differentiation of human embryonic stem cells into the gonadal fate. Overall, GeneCompass demonstrates the advantages of using artificial intelligence technology to decipher universal gene regulatory mechanisms and shows tremendous potential for accelerating the discovery of critical cell fate regulators and candidate drug targets.

DNA methylation plays multiple regulatory roles in crop development. However, the relationships of methylation polymorphisms with genetic polymorphisms, gene expression, and phenotypic variation in natural crop populations remain largely unknown. Here, we surveyed high-quality methylomes, transcriptomes, and genomes obtained from the 20-days-post-anthesis (DPA) cotton fibers of 207 accessions and extended the classical framework of population genetics to epigenetics. Over 287 million single methylation polymorphisms (SMPs) were identified, 100 times more than the number of single nucleotide polymorphisms (SNPs). These SMPs were significantly enriched in intragenic regions while depleted in transposable elements. Association analysis further identified a total of 5,426,782 cis-methylation quantitative trait loci (cis-meQTLs), 5078 cis-expression quantitative trait methylation (cis-eQTMs), and 9157 expression quantitative trait loci (eQTLs). Notably, 36.39% of cis-eQTM genes were not associated with genetic variation, indicating that a large number of SMPs associated with gene expression variation are independent of SNPs. In addition, out of the 1715 epigenetic loci associated with yield and fiber quality traits, only 36 (2.10%) were shared with genome-wide association study (GWAS) loci. The construction of multi-omics regulatory networks revealed 43 cis-eQTM genes potentially involved in fiber development, which cannot be identified by GWAS alone. Among these genes, the role of one encoding CBL-interacting protein kinase 10 in fiber length regulation was successfully validated through gene editing. Taken together, our findings prove that DNA methylation data can serve as an additional resource for breeding purposes and can offer opportunities to enhance and expedite the crop improvement process.