Computational modeling of genetic variations by multi-omics integration todecipher personal genome

多组学整合的遗传变异计算模型解读个人基因组 一个人的基因组通常包含数百万个遗传变异。通过评估了解这些变量 它们对一个人的表型的功能性影响，目前在人类遗传学和精确性方面引起了极大的兴趣。 药尽管全基因组关联研究（GWAS）或数量性状基因座（QTL）研究已经有 成功地鉴定出与性状或分子表型相关的变异，其中大多数是非编码的， 区域和连锁不平衡的阻碍，使偶然变异的识别和解释 难此外，这些发现大多是常见的变异，然而，罕见的和个体特异性的变异 在个人基因组中还未被充分研究。了解这些变异不仅可以解释缺失的遗传性 还能改善精准医疗。近年来，全基因组的出现和普及， 测序（WGS）和配对的多组学功能测定提供了一个前所未有的机会，以确定 罕见的和个体特异性的偶然变异。然而，大多数WGS研究的样本量是适度的 与GWAS相比，这使得WGS分析特别具有挑战性。然而，统计和 用于分析WGS的计算方法还不成熟。鉴于这些挑战和我独特的多- 学科培训，我的研究计划的总体目标是开发一种新型的机器学习， 统计和系统生物学方法识别，优先级和非编码的解释 通过整合GWAS、WGS和多组学功能测定， 通过识别用于疾病预防、诊断和治疗的个体化生物标志物。具体来说，在未来 五年后，我的实验室将（i）开发一种新的迁移学习方法来提高非编码的预测能力。 使用多维组学特征的偶然变体（ii）开发多组学集成综合扫描测试 改进从全基因组测序数据中鉴定罕见偶然变异的方法（iii）开发一个 个人基因组中非编码变异影响评分的综合计算框架（iv）开发一个 新的一类多性状方法，以提高表型预测使用全基因组遗传变异。 在 同时，在印第安纳州大学精准健康计划的支持下，我们将把这些方法应用于 来自印第安纳州阿尔茨海默病中心和印第安纳州多发性骨髓瘤生物样本库的不同研究， 科学发现。我们将与遗传学家和临床科学家密切合作， 发现。重要的是，我们将与实验室合作，以验证这些发现。符合我们之前 工作，我们将继续使所有开发的方法成为开源软件工具， 对生物医学研究界有用。

Computational modeling of genetic variations by multi-omics integration to decipher personal genome A person’s genome typically contains millions of genetic variants. Understanding these variants by assessing their functional impact on a person’s phenotype, is currently of great interest in human genetics and precision medicine. Though Genome-Wide Association Studies (GWAS) or Quantitative Trait Locus (QTL) studies have successfully identified variants associated with traits or molecular phenotypes, most of them are in noncoding regions and hampered by linkage disequilibrium, making the identification and interpretation of casual variants difficult. Moreover, most of these discoveries are common variants, however, rare and individual-specific variants in personal genome are underexplored. Understanding these variants will not only explain the missing heritability from GWAS but also improve the precision medicine. Recently, the advent and popularity of whole genome sequencing (WGS) and paired multi-omics functional assays provide an unprecedented opportunity to identify rare and individual-specific casual variants. However, the sample sizes of most WGS studies are modest compared to GWAS, making the WGS analysis particularly challenging. Nevertheless, statistical and computational methods for analyzing WGS are underdeveloped. Given these challenges and my unique multi- disciplinary training, the overall goals of my research program are to develop a novel class of machine learning, statistical and system biology approaches for the identification, prioritization and interpretation of noncoding variants by integrating GWAS, WGS and multi-omics functional assays, which will empower precision medicine by identifying individualized biomarkers for disease prevention, diagnosis and treatment. Specifically, in the next five years, my lab will (i) develop a novel transfer learning approach to improve the prediction of noncoding casual variants using multi-dimensional omics features (ii) develop a multi-omics integrated omnibus scan test to improve the identification of rare casual variants from whole-genome sequencing data (iii) develop an integrative computational framework for scoring impact of noncoding variants in personal genome (iv) develop a novel class of multi-trait methods to improve phenotype prediction using whole-genome genetic variations. In the meantime, supported by Indiana University Precision Health Initiative, we will apply the methodologies to different studies from Indiana Alzheimer’s Disease Center and Indiana Multiple Myeloma Biobank for novel scientific findings. We will work close with collaborated geneticists and clinician-scientists to interpret the discoveries. Importantly, we will work with experimental labs to validate the findings. In line with our previous work, we will continue to make all developed methods into open-source software tools that are accessible and useful to the biomedical research community.