PNNL

Abstract

We report the use of a novel high performance computing optimization method for the identification of proteins from unknown (environmental) samples. While computationally intensive compared to standard approaches, the optimization provides an effective way to control the false discovery rate for environmental samples and complements de novo peptide sequencing. Furthermore, the method can obviate the need to use DNA-based identification methods to find appropriate genomes when proteomic characterization is the primary goal and sub-species identification based on ribosomal phylogeny is not needed. We provide scaling and performance evaluations for the software that demonstrate the ability to carry out large-scale optimizations on 1258 genomes containing 4.2M proteins.