The genetic risk variants of polygenic diseases mainly reside in non-coding regions with their mechanisms largely elusive, hindering the diagnosis and treatment of these complex diseases. We have carried out mechanistic interpretation of polygenic diseases at multiple regulatory layers, including scRNA-seq and scATAC-seq (Cell 2023), m6A RNA methylation (Nature Genetics 2021) and H3K27ac histone modification (Nature Genetics 2023). Building upon these progresses, we further focus on deciphering the RNA-centric mechanisms of polygenic diseases using deep learning and statistical approaches.
In the first study, we present Translatomer, a multimodal transformer framework that predicts cell-type-specific Ribosome profiling from RNA-seq and gene sequence. We train Translatomer on 33 tissues and cell lines, and show that the inclusion of sequence substantially improves the prediction of ribosome profiling signal, indicating that Translatomer captures sequence-dependent translational regulatory information. Translatomer achieves accuracies of 0.72 to 0.80 for de novo prediction of cell-type-specific ribosome profiling. We develop an in silico mutagenesis tool to estimate mutational effects on translation and demonstrate that variants associated with translation regulation are evolutionarily constrained, both within the human population and across species. Notably, we identify cell-type-specific translational regulatory mechanisms independent of eQTLs for 3,041 non-coding and synonymous variants associated with complex diseases, including Alzheimer’s disease, schizophrenia, and congenital heart disease (Nature Machine Intelligence, 2024).
In the second study, we utilize quantitative trait loci (QTLs) as genetic instruments to delineate locus-specific directional maps of the crosstalk between m6A RNA methylation and two key epigenomic traits, including DNA methylation (DNAme) and H3K27ac. Using a combination of bidirectional Mendelian randomization (MR) and genetic colocalization approaches, we identify 47 loci indicative of the regulatory effects of m6A on H3K27ac (m6A-to-H3K27ac) in brain, as well as 4,733 m6A-to-DNAme loci in lung and muscle. In the reverse direction, we characterize 106 H3K27ac-to-m6A loci in brain and 61,775 DNAme-to- m6A in lung and muscle that represent the regulatory direction from epigenome to m6A. Notably, the loci in the ''DNAme-to- m6A'' direction show strong enrichment in enhancers. We prioritize 1,767 genetic variants that may modulate the crosstalk between m6A and DNA methylation for asthma and expiratory flow traits in lung, as well as 249 for coronary artery disease, blood pressure, and pulse rate in muscle. This integrative analysis establishes comprehensive regulatory paths for disease variants, such as the rs3768410-DNAme- m6A-asthma and the rs56104944-m6A-DNAme-hypertension axes (Cell Genomics, 2024).
These studies together provide molecular targets with a mechanistic foundation towards prediction, molecular diagnosis, and precision therapy for human polygenic diseases.