Advanced machine learning models to integrate multi-modal biomedical datasets for gene regulation and precision medicine

University Of North Texas, Denton TX

Investigators

Linked publications & trials

Abstract

Vast amounts of multi-modal biomedical datasets have become available due to advances in high-throughput biomedical technologies. Each of these distinct data modalities such as genomics, epigenomics, and transcriptomics (collectively called as “multi-omics”) offers complementary information about the underlying biology of living systems. Therefore, there is an urgent need for scalable methods capable of integrating multi-modal datasets across millions of individuals. Our research program is devoted to the development of open-source integrative computational tools designed to analyze high-dimensional multi-modal biomedical datasets such as multi-omics and electronic health records (EHR) data. In this MIRA renewal application, we aim to develop generalizable, biologically inspired, and interpretable machine learning solutions to address some of the key challenges present in the biomedical datasets such as data irregularities, dependencies between data modalities, and missing and noisy data. We will develop novel computational methods based on machine learning, deep learning, and graph representation learning to integrate multi-modal biomedical datasets to build generalizable and interpretable models. These models will be used for various prediction tasks such as predicting clinical outcomes, inferring regulatory networks, and identifying drugs that can be repurposed for other diseases. The vision of our research program is to develop open-source computational tools that efficiently integrate multi-modal biomedical datasets, enhancing our understanding of gene regulatory interactions and disease mechanisms. By leveraging machine learning, deep learning, graph representation learning, and foundational models, we aim to advance precision medicine, facilitating more informed treatment decisions.