CAREER: Brain Imaging Genetics via multimodal modular structure querying

University Of Pittsburgh, Pittsburgh PA

Investigators

Abstract

Brain imaging genetics, which integrates the merits of brain imaging technologies and genetic data, has the potential to improve our understanding of the human brain. Recent studies have shown that brain imaging genetics is a powerful tool to discover the polygenic contributions for brain disorders and quantitatively characterize the neural systems affected by risk gene variants. This project will develop a series of novel computational tools to address the critical challenges and bottlenecks in current brain imaging genetics research, and will directly impact biomedical informatics, brain research, and data science. The success of this project will be used to develop new curriculums that incorporate research into the classroom and provide students from under-represented groups with opportunities to participate in biomedical and machine learning research. The main challenges in current brain imaging genetics are as follows. First, most existing brain imaging genetics studies assume the linear relationship between genes and imaging features. Considering the high dimensionality of brain magnetic resonance imaging (MRI) data and genetic data, this linearity is too simplistic. Second, traditional brain MRI research is suboptimal in characterizing brain dynamics because they usually focus on scalar statistics, which reduce the complex brain imaging data to a one-dimension and discard important informative brain network structures. In this project, we choose the brain modular structure as the feature representations. These kinds of representations can better describe the intermediate scale of brain network organization, rather than any global or local scales. The brain modular structure provides a promising bridge as the intermediate neuroendophenotype with a smaller dimension and a more focused objective to link genotypic and phenotypic traits. Moreover, how to derive the modular structure from multimodal data has not been well addressed. This project will provide efficient and biologically meaningful tools to map polygenetic components to phenotypes with the aid of brain modular features. The successful development of these new tools will have an immediate and strong impact on brain research, network science, and machine learning. Moreover, this project offers multidisciplinary training opportunities for trainees at all levels from K-12 to postdoctoral levels. Outcomes will be openly disseminated in peer-reviewed articles, outreach programs, and in the form of code/data repositories to maximize impact. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.