Linking genetics to cellular behavior and disease via multimodal data integration

点击翻译按钮获取中文摘要

Project Summary / Abstract Studies of mechanisms underlying genetic associations with disease have primarily focused on gene regulatory mechanisms. In contrast, the impact of genetic variation on cellular phenotypes such as morphology and behavior is poorly understood, despite their importance in disease progression. This poor understanding is in part because measurement of some cellular phenotypes such as electrophysiological response patterns require skilled manual labor and is performed cell by cell, and is thus low throughput. Furthermore, some cellular phenotype assays require live cells, which for cell types such as neurons are prohibitively challenging to obtain from humans. The goal of this proposal is to develop deep learning-based frameworks for characterizing how molecular and cellular phenotypes covary using multimodal datasets, then to predict how these cellular behaviors mediate the effect of genetic variation on the risk of illnesses such as psychiatric and neurodevelopmental disorders. We will achieve this overall goal through the development of three frameworks. Multimodal models linking molecular and cellular phenotypes of cells. By linking gene regulation with cellular phenotypes such as neuron electrophysiology and morphology, we can then understand how changes at the molecular level propagate to cellular phenotypes and vice versa. Furthermore, we can use these models to impute cellular phenotypes when they cannot be measured experimentally. Identification of cellular phenotypes that mediate genetic risk of mental disorders. We will jointly model genotype, gene expression, cellular phenotypes and disease risk to generate mechanistic hypotheses about the mediation of genetic effects on disease risk through cellular phenotypes. Prediction of cellular phenotypes associated with disease progression. We will develop a prediction framework for exploring which cellular phenotypes change significantly with disease progression. By applying our imputation framework developed above, we will predict changes in neuron electrophysiological response and morphology associated with a range of psychiatric and neurodevelopmental disorders. We expect completion of this project to yield generalizable computational frameworks for linking genetics to molecular and cellular phenotypes for diverse cell types and organs.