Integrative Analysis of High-Throughput Toxicity Screening Data

• Activity Number: 287
• Type: Invited
• Date/Time: Tuesday, August 2, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Section on Risk Analysis
• Author(s): Eric F. Lock*
• Companies: University of Minnesota
• Keywords: Data integration ; Toxicity screening ; Matrix factorization ; Multi-way arrays ; Imputation

Abstract

We describe an exploratory analysis of an in vitro toxicity screening experiment, using linked matrix factorizations. These data include measures of toxicity from 884 lymphoblast cell lines from different individuals for 176 chemical compounds, genomic data for each cell line, and high-dimensional attribute data for each chemical. Thus, our data takes the form of three linked arrays: toxicity (individuals X chemicals), genomic (individuals X genes) and attributes (attributes X chemicals). New methods for the simultaneous factorization of linked data arrays reveal genomic and chemical attribute patterns that account for inter-individual variability in toxicological risk.