Integrating Common Fund Program Datasets for Feature Selection and Machine Learning

Dr. Deanne M. Taylor will present a graph database that enables users to search and grow relationships between genes, variants, and pathways on relevant genomic locations of interest. This graph database is modeled on structures in NIH’s Unified Medical Language System (UMLS) and comprises an integration of biological knowledge extracted from five NIH Common Fund data sets. Graph database users can query for select features to study new variant-to-variant and gene-to-gene relationships in genetic cohorts and determine other types of complex relationships among entities in NIH’s Common Fund data sets and UMLS annotations.

Data are derived from five Common Fund data sets:

• Knockout Mouse Phenotyping Program 
• Genotype-Tissue Expression 
• The Human BioMolecular Atlas Program
• Gabriella Miller Kids First Pediatric Research Program
• Library of Integrated Network-based Cellular Signatures