Premium
Refining spectral library searching
Author(s) -
Rudnick Paul A.
Publication year - 2013
Publication title -
proteomics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.26
H-Index - 167
eISSN - 1615-9861
pISSN - 1615-9853
DOI - 10.1002/pmic.201300426
Subject(s) - computer science , information retrieval , false discovery rate , similarity (geometry) , linear discriminant analysis , ground truth , sensitivity (control systems) , sequence (biology) , data mining , machine learning , artificial intelligence , engineering , chemistry , genetics , electronic engineering , biology , image (mathematics) , gene , biochemistry
Spectral library searching has many advantages over sequence database searching, yet it has not been widely adopted. One possible reason for this is that users are unsure exactly how to interpret the similarity scores (e.g., “dot products” are not probability‐based scores). Methods to create decoys have been proposed, but, as developers caution, may produce proxies that are not equivalent to reversed sequences. In this issue, Shao et al. ( Proteomics 2013, 13 , 3273–3283) report advances in spectral library searching where the focus is not on improving the performance of their search engine, SpectraST, but is instead on improving the statistical meaningfulness of its discriminant score and removing the need for decoys. The results in their paper indicate that by “standardizing” the input and library spectra, sensitivity is not lost but is, surprisingly, gained. Their tests also show that false discovery rate (FDR) estimates, derived from their new score, track better with “ground truth” than decoy searching. It is possible that their work strikes a good balance between the theory of library searching and its application. And as such, they hope to have removed a major entrance barrier for some researchers previously unwilling to try library searching.