ReCom: A semi-supervised approach to ultra-tolerant database search for improved identification of modified peptides

Open-search methods allow unbiased, high-throughput identification of post-translational modifications in proteins at an unprecedented scale. The performance of current open-search algorithms is diminished by experimental errors in the determination of the precursor peptide mass. In this work we propose a semi-supervised open search approach, called ReCom, that minimizes this effect by taking advantage of a priori known information from a reference database, such as Unimod or a database provided by the user. We present a proof-of-concept study using Comet-ReCom, an improved version of Comet-PTM. Comet-ReCom increased identification performance of Comet-PTM by 68%. This increased performance of Comet-ReCom to score the MS/MS spectrum comes in parallel with a significantly better assignation of the monoisotopic peak of the precursor peptide in the MS spectrum, even in cases of peptide coelution. Our data demonstrate that open searches using ultra-tolerant mass windows can benefit from using a semi-supervised approach that takes advantage from previous knowledge on the nature of protein modifications. For Table of Contents Only O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=106 SRC="FIGDIR/small/535358v1_ufig1.gif" ALT="Figure 1"> View larger version (31K): org.highwire.dtl.DTLVardef@1069b1borg.highwire.dtl.DTLVardef@b12bfdorg.highwire.dtl.DTLVardef@1bf23b6org.highwire.dtl.DTLVardef@1c1c02b_HPS_FORMAT_FIGEXP M_FIG C_FIG