Quality Assessment of MS/MS Spectra Using Variable Selection and Support Vector Machine

摘要

High-throughput proteomics experiments produce large amounts of MS/MS data, but many are of too low quality to be utilized. Filtering out the low quality MS/MS spectra is one of the strategies to increase computational speed of database searching. We investigate the variables proposed in previous literatures, and collect 63 features for our study, including 8 new variables proposed in this paper. MRMR is used to select important variable set for 4 kinds of mass spectrometer platforms: Thermo LTQ-FT, LTQ LCQ, and Waters/Micromass QTOF. A Support Vector Machine based method is used to assess the quality of MS/MS spectra. Several datasets from different data sources and different mass spectrometers are applied to test the accuracy and the reproducibility of our method. In order to prove the capability of our method, we compare it with msmsEval on the test datasets, and the results show that our method achieves higher accuracy and better performance than msmsEval. The programs can be obtained from the authors by request.