Integrative predictive medicine to identify disease causes, develop cures, and optimize patient care

Project Summary/Abstract The Laboratory for Pathology Dynamics was founded on the principle that pathology is multifactorial, multi- scalar, and driven by complex interacting dynamics. This proposal builds and applies new predictive medicine technology that enables multifactorial, multi-scalar data integration to identify disease causes, develop cures, and optimize patient care. Key knowledge gaps addressed include: 1.) integration of siloed relationships for interactive literature-based discovery across 33+ million journal articles, including the development of a comprehensive biomedical knowledge graph software called SemBioSys; 2) high-throughput automated meta- analysis to quantify effect size across cohorts and scales using a combination of newly developed deep learning, active learning, interactive weak supervision and human in the loop artificial intelligence (AI) algorithms into a single interactive software; 3) de novo multimodal machine learning algorithms to investigate multifactorial, multi-scalar disease dynamics; 4) human in the loop methods for enhancing AI accuracy, utility, and interpretability, which leverages our large test bed of diverse high school and undergraduate curators acting as subject matter experts. The developed technology will be applied to a bevy of multifactorial diseases in collaboration with clinicians and experimentalists: neurological diseases (Alzheimer’s Disease, Amyotrophic Lateral Sclerosis, Parkinson’s Disease), cancers (adult and pediatric acute and chronic leukemias and solid tumor cancers), cardiovascular disease, and infectious disease. The significance of this project is integrative technology to identify and rank novel disease mechanisms, discover and prioritize novel and repurposed drugs, and to optimally stratify patients for fair and equitable clinical trials that minimize health outcome disparities. The innovation includes new integrative biomedical natural language processing (NLP), multimodal machine learning, and dynamic disease modeling and event sequencing algorithms that can be generalized to all of medicine.