Building models and tools to integrate multiscale brain imaging and multi-omics data

Title: Building models and tools to integrate multiscale brain imaging and multi-omics data Project Summary Our recent work on integrative analysis of multiscale brain imaging and multi-omics shows added value in disease subtyping and biomarker detection. However, the following significant challenges remain: (1) complex interactions with linear and nonlinear features occur at multiple scales (e.g., interactomes) but are rarely incorporated into multi-modal analyses, (2) existing integrative approaches combine multi-modal brain imaging with multi-omics but exclude phenotypic traits and a priori biological knowledge, and (3) current predictive models are not easily interpretable. Without models that combine complementary imaging and omics information to accurately predict behavioral phenotypes and identify the key parameters and biomarkers that govern them, we are unlikely to increase accuracy of disease risk prediction and improve patient outcomes with the determination of the underlying biological mechanisms. In the proposed research, we will address these challenges by (i) developing innovative methods and tools for brain imaging, omics, and behavioral data integration and (ii) validating them in the context of mental health. Our application is the clinically significant problem of defining and predicting multiple psychiatric disorders— currently difficult, at best—with the driving vision of advancing brain research toward precision psychiatry. Our multidisciplinary team with a decade of productive collaborations will address the following Specific Aims: (1) Develop nonlinear modeling approaches to detect complex interaction networks within and across scales; (2) Develop and validate phenotype-guided integrative approach to combine multiscale imaging and omics data while incorporating biological knowledge; (3) Build an interpretable predictive model linking multiscale brain imaging and omics data, including their complex interactions, with behavioral phenotype; and (4) Disseminate multimodal data interaction and integration tools to the neuroimaging research community. The proposed research will advance the field of neuroimaging by filling the gap in interpretable multimodal data integration approaches for comprehensive brain imaging and multi-omics data analysis. The methods and tools developed by this project will be used to predict clinical outcomes and classify psychiatric disorders via their underlying biological mechanisms. By disseminating the tools and software to the research community, this work will have a broad and sustained impact on individualized disease diagnosis and treatment.