NEURAL NETWORK AS AN APPROACH TO THE CLASSIFICATION OF UNASSIGNED LC-MS DATA IN PROTEOMETABOLOMICS

Abstract

Proteomic (MASCOT, Andromeda, OMSSA, etc.) and metabolomic (Scripps, AMDIS, etc.) search algorithms for identification are limited to reference libraries and, as a result, a large amount of data is lost (modified proteins, isoforms). Bypassing the identification stage, you can significantly increase the amount of data for classification tasks. This makes it possible to successfully cluster the studied classes at the cost of information on molecular categorical predictors.