Deep Learning Based Genetic Risk Prediction for Type 1 Diabetes

项目摘要 I型糖尿病(T1D)是一种儿童自身免疫性疾病，由遗传和 环境因素。在具有T1D遗传易感性的一部分个体中，环境触发因素 激发自身免疫反应，以胰岛为靶点并破坏，导致糖尿病前期 最终导致糖尿病。T1D预防研究中的一个关键障碍是识别和直接招募儿童 具有将T1D发展为预防试验的强烈遗传易感性。稳健的遗传风险评分 (GRS)将允许识别T1D高危儿童，将他们招募到T1D预防中 试验，以及随后对新干预措施的测试。我的目标是1)优化多层前馈神经网络 可用于直接将新生儿纳入T1D预防试验的网络遗传风险预测指标；以及2) 确定推测的、新的导致T1D的SNP及其相互作用。目标1的完成将提供一个 为T1D研究社区提供更好的GRS，可用于识别具有较高遗传风险的儿童 增加了未来T1D预防临床试验的统计能力。完成 AIM 2将提供对T1D发育的分子驱动因素的更深入的生物学理解，以及 T1D预防试验潜在的新治疗靶点。这个项目的成功完成将既 帮助了解1型糖尿病的遗传原因，并帮助预防这种疾病。

Project Summary Type I diabetes (T1D) is an autoimmune disease of childhood caused by a combination of genetic and environmental factors. In a subset of individuals with a genetic predisposition to T1D, environmental triggers instigate an autoimmune response which targets and damages pancreatic islets, leading to pre-diabetes and ultimately diabetes. A critical barrier in T1D prevention research is to identify and directly enroll children with a strong genetic predisposition for developing T1D into prevention trials. A robust genetic risk score (GRS) would allow for the identification of children at high-risk of T1D, their recruitment into T1D prevention trials, and subsequent testing of novel interventions. I aim to 1) optimize a multi-layer feedforward neural network genetic risk predictor that can be used to enroll newborns directly into T1D prevention trials; and 2) identify putative, novel T1D-causing SNPs, and their interactions. Completion of aim 1 would provide a better GRS to the T1D research community, which can be used to identify children with higher genetic risk of T1D development, increasing the statistical power of future T1D prevention clinical trials. Completion of aim 2 will provide a deeper biological understanding of the molecular drivers of T1D development, and potential new therapeutic targets for T1D prevention trials. Successful completion of this project will both help understand the genetic causes of type 1 diabetes and help prevent the disease.