pathQTL: Integrative Multi-Omics Causal Inference of Molecular Mechanisms Leading to Neuropsychiatric Illness

项目摘要 影响神经精神障碍风险的多种常见基因变异(例如，精神分裂症， 严重抑郁障碍和阿尔茨海默病)最近被识别和复制，提供了一种 以此为立足点来研究这些疾病的原因。神经精神病学遗传学的关键下一步是从 了解这种遗传变异如何影响分子、细胞和 大脑的循环，导致复杂的紊乱。许多数据集，包括那些由我们自己的实验室生成的数据集， 在多个生物学水平上建立了基因和人脑特征之间的直接联系(分子： 染色质可及性，表达；细胞：形态；回路：大脑皮层结构)，称为数量性状 座位(QTL)。在这里，我们将整合多个生物学水平的QTL，以便在统计上区分优先顺序 基因变异造成复杂神经精神疾病风险的因果途径。因果关系 建模远远超越了以前的协同本地化工作，因为它允许对昂贵的功能 对引起疾病的细胞或分子变化的验证实验，而不是那些 是无序的结果或独立于无序。它还允许对关键实验进行推断 参数包括细胞类型和发育时间。最后，因果推论结合在一起时 多个水平的生物学和多个紊乱风险位点允许评估在生物学 水平、细胞类型或发育时间，这是治疗靶向的关键信息。我们会 利用贝叶斯概率网络和因果推理的计算和统计框架 在利用关联摘要统计信息的新框架中，以及在 同样的个体几乎总是不可行的。随后，我们将对该分子进行实验验证 在原代人类神经前体细胞中使用表观遗传工程预测我们的模型，并反过来 修改计算模型。排列紊乱相关变异的原因分子途径的优先顺序，以及 识别相关的细胞类型和发育阶段将提高验证的成功率 实验，并以公正的方式阐明神经精神障碍的机制。

Project Summary A multitude of common genetic variants influencing risk for neuropsychiatric disorders (e.g., schizophrenia, major depressive disorder, and Alzheimer’s disease) have recently been identified and replicated, providing a foothold into the causes of these disorders. The critical next step in neuropsychiatric genetics is to move from a risk locus in the genome to an understanding of how this genetic variation influences molecules, cells, and circuits of the brain, leading to complex disorders. Many datasets, including those generated by our own labs, have established direct links between genotype and human brain traits at multiple levels of biology (molecular: chromatin accessibility, expression; cellular: morphology; circuit: gross brain structure), termed quantitative trait loci (QTLs). Here, we will integrate QTLs across multiple levels of biology in order to statistically prioritize causal pathways by which genetic variation creates risk for complex neuropsychiatric disorders. Causal modeling goes well beyond previous co-localization work, as it allows the prioritization of expensive functional validation experiments for cellular or molecular changes that are a cause of the disorder, rather than those that are a consequence or independent of the disorder. It additionally allows inference of key experimental parameters including cell-type and developmental time period. Finally, causal inference when combined across multiple levels of biology and multiple disorder risk loci allows for assessment of convergence at a biological level, cell-type, or developmental time period, which is critical information for therapeutic targeting. We will leverage the computational and statistical frameworks of Bayesian probabilistic networks and causal inference in a new framework that utilizes association summary statistics, as well-powered multi-level data collected on the same individuals is almost always infeasible. Subsequently, we will experimentally validate the molecular predictions of our model using epigenetic engineering in primary human neural progenitor cells, and in turn revising the computational models. Prioritizing causal molecular pathways of disorder associated variants, and identifying the relevant cell-type and developmental stage will increase the success rate of validation experiments and shed light on mechanisms of neuropsychiatric disorders in an unbiased manner.