复杂疾病致病调控变异及其靶基因预测方法研究

Precise identification disease-causal regulatory variant in the non-coding genome will significantly facilitate the understanding of genetic mechanism for complex diseases. Relying on particular cellular environment, causal regulatory variant alters the expression pattern of targeted gene and promotes the disease development. We previously developed a series of computational methods to predict and prioritize non-coding functional regulatory variants, however, these methods either lack the abilities to identify causal regulatory variant under particular disease condition, or cannot link the regulatory variant to its gene targets at satisfactory resolution. By integrating large-scale tissue/cell type-specific epigenomes and 3D genomes data, we will design several statistical models to infer causal tissue/cell type and causal regulatory variant for particular disease from summary statistics of genome-wide association study, and predict targeted genes of regulatory variant with high resolution using attention model-based deep learning network. We will experimentally evaluate some of our predictions by in vitro assays. This project will provide novel methods to fine-map disease-causal regulatory variant and associate targeted genes at high resolution, which promotes the in-depth investigation of pathogenesis mechanisms for complex diseases.

准确地鉴定基因组非编码区域的致病调控变异及其作用靶基因对理解复杂疾病遗传机制具有重要意义。致病调控变异依赖于特定的细胞环境，通过多种机制影响靶基因表达模式，进而导致疾病的发生发展。我们前期开发了一系列生物信息学方法预测和排优化非编码功能调控变异，然而，目前的计算方法仍缺乏在特定的复杂疾病条件下准确的预测因果致病调控变异的能力，也无法在精细的粒度上关联调控变异到其相互作用靶位点。本课题拟利用大规模组织细胞特异的表观基因组和三维基因组数据，设计统计计算模型从全基因组关联分析汇总统计结果中推断疾病因果组织/细胞和致病调控变异；构建基于注意力模型的深度学习网络在高分辨率水平上预测调控变异的靶基因；并通过体外实验验证若干预测结果。本课题将为精确定位复杂疾病致病调控变异及其靶基因提供了新的计算方法，促进深入理解复杂疾病遗传致病机制。