Critiquing Bioinformatics Resources When ID Mapping Is Needed to Integrate Genomic and Proteomic Studies

• Activity Number: 581
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Roger Day*+ and Kevin McDade and Uma Chandran and Alex Lisovich
• Companies: University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh
• Keywords: bioinformatics identifiers ; mixture model ; genomics ; proteomics ; correlation bootstrapping

Abstract

Microarray gene expression (mRNA) studies and tandem mass spectrometry (MS/MS) proteomics studies each reflect protein abundance, but imperfectly for different reasons. Integrated analyses require correct matching between protein identifiers and microarray probeset identifiers. With 98 samples from an endometrial cancer study, MS/MS yielded 12288 UniProt accessions (ACs). We retrieved matches from ACs to probesets ID's from three popular bioinformatics web services. The services displayed astonishing discordance. For each proposed match, we could merge and correlate the MS/MS and mRNA data. The proteome-genome correlations support a lovely mixture model. Bootstrap-weighted regression analyses then evaluate the quality of identifier matches, presuming that better matching should usually generate stronger assay correlations. Here we report which service did the best-and the worst.