肝癌是我国最常见的恶性肿瘤之一，其病因的多元性提示了肝癌基因组的高度异质性和复杂性。国内外针对肝癌基因组的研究主要聚焦在对编码区突变的解读，虽然非编码区的序列变异约占到已知变异的绝大多数，但对非编码变异的功能解读，特别是其与肿瘤之间关系的研究才刚刚起步。本项目拟整合大量的一维表观基因组学数据和三维染色质结构数据以及大规模表达数量性状基因座（eQTL）数据，利用深度神经网络算法建立功能非编码变异对靶基因表达水平影响的定量预测模型。其次，将模型应用于肝癌肿瘤基因组测序数据中，鉴定一系列与肝癌相关的高频非编码变异。最后，在肝癌细胞系中利用基因组编辑技术（CRISPR）对预测得到的高频功能非编码变异和表达调控关系进行验证。

Liver cancer is one of the most fatal malignant cancers in China, and its various causing factors suggest the complexity and heterogeneity in the underlying genome. Most studies both home and abroad mainly focused on interpreting the variants in the coding regions, although noncoding variants take up the vast majorities of currently known variants, the functions of noncoding variants, especially their relationships with cancers are still largely unknown. In this project, we will integrate a large amount of one dimensional epigenomics data and three dimensional genome architecture data, and expression quantitative trait loci (eQTL), and will develop a deep neural network model to identify functional noncoding variants and to predict their effects on target genes. We will then apply this model to the whole genome sequencing data of liver cancer to detect a set of recurrent noncoding variants associated to this cancer. At last, we will experimentally validate these recurrent noncoding variants and their effects on target genes in HepG2 cell line using CRISPR genome editing technology.