Augmenting the CMDKP with advanced features for understanding biological mechanisms

Broad Institute, Inc., Cambridge MA

Investigators

Linked publications & trials

Abstract

Title: Augmenting the CMDKP with advanced features for understanding biological mechanisms Project Summary The Common Metabolic Diseases Knowledge Portal (CMDKP) aggregates genetic, genomic, and phenotypic data to accelerate discovery of therapies for type 2 diabetes and related common metabolic disorders. Under the parent UM1 grant we have (a) identified and aggregated datasets for inclusion in the portal; (b) implemented bioinformatic pipelines for processing these data through methods for predicting the effects of GWAS associations; and (c) built interfaces to access an integrated resource of these datasets and method results. This work has produced a widely used knowledge portal and made important advances in increasing genetic and genomic data accessibility. To increase the impact of the portal on the ultimate AMP-CMD goal of informing on all major questions regarding genetically driven target identification and validation, in this supplement we will make higher-level queries about common metabolic disease mechanisms easier to address through the CMDKP. We will address this goal in three ways. First, we will extend and apply statistical methods to hierarchically extract and organize information from genetic and genomic datasets within the CMDKP. These methods will allow us both to present higher-level relationships among biomedical entities that can be more easily interpreted and queried by users, and it will also allow users to more easily organize and locate functionality and datasets within CMDKP. Second, we will improve the user interface of the CMDKP to make it easier for users to conduct integrative queries of its functionality. We will achieve this by taking advantage of new AI technologies for large language models to redesign the CMDKP user interface, and we will also make the results of the CMDKP available as a knowledge graph and open API catalogue that large-language models can invoke via function calling. Finally, we will improve user access to the new results and functionality within the CMDKP through new outreach and training approaches. These new approaches will target CMDKP data contributors, method developers, and users, taking advantage of new AI technologies that allow us to tailor training to each user’s background. Collectively, this work will significantly extend the CMDKP with the next generation of tools to use genetic and genomic data to understand common metabolic disease mechanisms and ultimately develop new treatments.