Significance: Chromatin plays a critical role in the regulation of gene expression. Interactions among chromatin regulators, sequence-specific transcription factors, and cis -regulatory sequence elements are the main driving forces shaping context-specific chromatin structure and gene expression. However, because of the large number of such interactions, direct data on them are often missing in most cellular contexts. The purpose of the present work is to show that, by modeling matched expression and accessibility data across diverse cellular contexts, it is possible to recover a significant portion of the information in the missing data on binding locations and chromatin states and to achieve accurate inference of gene regulatory relations.

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.

Computational inference of novel therapeutic values for existing drugs, i.e., drug repositioning, offers the great prospect for faster and low-risk drug development. Previous researches have indicated that chemical structures, target proteins, and side-effects could provide rich information in drug similarity assessment and further disease similarity. However, each single data source is important in its own way and data integration holds the great promise to reposition drug more accurately. Here, we propose a new method for drug repositioning, PreDR (Predict Drug Repositioning), to integrate molecular structure, molecular activity, and phenotype data. Specifically, we characterize drug by profiling in chemical structure, target protein, and side-effects space, and define a kernel function to correlate drugs with diseases. Then we train a support vector machine (SVM) to computationally predict novel drug-disease interactions. PreDR is validated on a well-established drug-disease network with 1,933 interactions among 593 drugs and 313 diseases. By cross-validation, we find that chemical structure, drug target, and side-effects information are all predictive for drug-disease relationships. More experimentally observed drug-disease interactions can be revealed by integrating these three data sources. Comparison with existing methods demonstrates that PreDR is competitive both in accuracy and coverage. Follow-up database search and pathway analysis indicate that our new predictions are worthy of further experimental validation. Particularly several novel predictions are supported by clinical trials databases and this shows the significant prospects of PreDR in future drug treatment. In conclusion, our new method, PreDR, can serve as a useful tool in drug discovery to efficiently identify novel drug-disease interactions. In addition, our heterogeneous data integration framework can be applied to other problems.

High-altitude adaptation of Tibetans represents a remarkable case of natural selection during recent human evolution. Previous genome-wide scans found many non-coding variants under selection, suggesting a pressing need to understand the functional role of non-coding regulatory elements (REs). Here, we generate time courses of paired ATAC-seq and RNA-seq data on cultured HUVECs under hypoxic and normoxic conditions. We further develop a variant interpretation methodology (vPECA) to identify active selected REs (ASREs) and associated regulatory network. We discover three causal SNPs of EPAS1, the key adaptive gene for Tibetans. These SNPs decrease the accessibility of ASREs with weakened binding strength of relevant TFs, and cooperatively down-regulate EPAS1 expression. We further construct the downstream network of EPAS1, elucidating its roles in hypoxic response and angiogenesis. Collectively, we provide a systematic approach to interpret phenotype-associated noncoding variants in proper cell types and relevant dynamic conditions, to model their impact on gene regulation.

Chromatin accessibility plays an essential role in controlling cellular identity and the therapeutic response of human cancers. However, the chromatin accessibility landscape and gene regulatory network of pancreatic cancer are largely uncharacterized. Here, we integrate the chromatin accessibility profiles of 84 pancreatic cancer organoid lines with whole-genome sequencing data, transcriptomic sequencing data and the results of drug sensitivity analysis of 283 epigenetic-related chemicals and 5 chemotherapeutic drugs. We identify distinct transcription factors that distinguish molecular subtypes of pancreatic cancer, predict numerous chromatin accessibility peaks associated with gene regulatory networks, discover regulatory noncoding mutations with potential as cancer drivers, and reveal the chromatin accessibility signatures associated with drug sensitivity. These results not only provide the chromatin accessibility atlas of pancreatic cancer but also suggest a systematic approach to comprehensively understand the gene regulatory network of pancreatic cancer in order to advance diagnosis and potential personalized medicine applications.

Background: In general, gene function prediction can be formalized as a classification problem based on machine learning technique. Usually, both labeled positive and negative samples are needed to train the classifier. For the problem of gene function prediction, however, the available information is only about positive samples. In other words, we know which genes have the function of interested, while it is generally unclear which genes do not have the function, i.e. the negative samples. If all the genes outside of the target functional family are seen as negative samples, the imbalanced problem will arise because there are only a relatively small number of genes annotated in each family. Furthermore, the classifier may be degraded by the false negatives in the heuristically generated negative samples. Results: In this paper, we present a new technique, namely Annotating Genes with Positive Samples (AGPS), for defining negative samples in gene function prediction. With the defined negative samples, it is straightforward to predict the functions of unknown genes. In addition, the AGPS algorithm is able to integrate various kinds of data sources to predict gene functions in a reliable and accurate manner. With the one-class and two-class Support Vector Machines as the core learning algorithm, the AGPS algorithm shows good performances for function prediction on yeast genes. Conclusion: We proposed a new method for defining negative samples in gene function prediction. Experimental results on yeast genes show that AGPS yields good performances on both training and test sets. In addition, the overlapping between prediction results and GO annotations on unknown genes also demonstrates the effectiveness of the proposed method.