ALZHEIMERS RESEARCH PROJECT: Multi-omic Knowledge Transfer for Target Discovery and Due Diligence

National Institute On Aging

Investigators

Abstract

This project represents a comprehensive initiative to revolutionize drug target discovery in Alzheimer's Disease and related dementias (AD/ADRD) through the development of omicSynth, an innovative multi-omic resource that integrates large-scale genetic and genomic data with artificial intelligence tools. The project has successfully achieved several major milestones, including the publication of groundbreaking research identifying over 150 genetic targets associated with neurodegenerative diseases, the development of publicly accessible web platforms, and significant contributions to NIH-wide AI initiatives. The team has demonstrated exceptional productivity with multiple publications, tool releases, and collaborative efforts that position this work at the forefront of precision medicine in neurodegeneration research. The project's impact extends beyond traditional research boundaries by incorporating cutting-edge AI technologies, including large language models and knowledge graphs, to create next-generation tools for drug discovery and repurposing. The successful integration of multi-omic data types, disease genome-wide association study summary statistics, and drug information has created a robust platform that democratizes access to complex genomic data and accelerates the translation from genetic discoveries to therapeutic interventions. This project aims to develop an open multi-omic resource to identify druggable targets in Alzheimer's Disease and Alzheimer's Disease related dementias (AD/ADRD) using large-scale genetic and genomic data. Given the importance of anchoring therapeutic targets to a disease mechanism substantiated by genetic evidence, we developed omicSynth: a dynamic, open, and accessible resource that leverages large-scale genetic and genomic data for the identification of therapeutic targets in the AD/ADRD space. We prioritized identified genes as therapeutic targets of interest based upon known small-molecule druggability and product market information. This has grown into a next generation work in progress slated for Q4 2025 integrating network community mapping, knowledge graphs and modern large Language Models to discovery and carry out *-omics grounded due diligence on drug targets. This will include a complimentary public facing tool in the spirit of omicSynth that incorporates multi-omic data, disease genome wide association study summary statistics, drug data, and other relevant data types, eliminating barriers to drug discovery and drug repurposing and potentially enabling precision medicine in the AD/ADRD space. Using multi-omics integration methods, AI-assistance, and most importantly, community input to better parse and interpret the data presented by the platform, we aim to make our community resource a robust tool for AD/ADRD research. We recently published a study in the American Journal of Human Genetics that used summary-data-based Mendelian Randomization to identify over 150 genetic targets associated with neurodegenerative diseases, including 41 novel targets. A companion preprint relating to the new integration and vetting of AI in drug discovery and AD/ADRD research has been submitted and is under a third revision at Lancet Digital Health [https://pmc.ncbi.nlm.nih.gov/articles/PMC11760394/]. Some of this code was contributed to the larger NIH-wide ChiIRP AI tooling [https://irp.nih.gov/catalyst/33/2/chirp-a-chatgpt-model-for-the-nih-intramural-community] as part of our collaborative efforts with other teams across NIH. Data and resource dissemination plan will comply with FAIR data standards. Results have been published in a peer-reviewed journal (PMID: 38181731), and code deposited in the relevant CARD Github and Zenodo. We also developed a web platform (a streamlit app) to explore therapeutic targets: https://nih-card-ndd-smr-home-syboky.streamlit.app and an internal AI tool for annotation, information synthesis and summary statistic manipulation called CARD.AI. This will leverage integration of our CRISPR knowledge platforms and collaborations with the Broad Institute's Neurodegenerative Disease Knowledge Portal. A new version of omicSynth will be integrated into CARD.AI in 2025 and a publication in late 2025 or early 2026.