Peptide Mass Fingerprinting (PMF) plays an irreplaceable role in nowadays tandem proteomics due to its higher sample
throughput, higher level of specificity for single peptides and lower level of sensitivity to unsuspected post-translational
modifications compared to MS/MS. The PMF method would become more attractive if we could improve the accuracy of protein
identification. With this motivation, we have proposed and evaluated a feature-matching based uniform approach using support
vector machines (SVMs) to incorporate individual concepts and conclusions for accurate PMF. The SVMs approach focused
on the inherent attributes and critical issues of theoretical spectrum (peptides), experimental spectrum (peaks) and spectrum
(masses) alignment. The experimental peak intensity was introduced to the algorithm. An optimal SVMs model with 491 out
of 35,640 feature-matching patterns outperformed Mascot, MS-Fit, ProFound and Aldente with a high-performance evaluation
on a standard PMF set of 225 items. Now, the approach is extended with a web server, FMP, to identify protein from MS1 data.
The web implementation contains several features: (i) a local secondary database PUD (Peptide Uniqueness Database) have been
constructed to provide each theoretical peptide?s sequence and mass uniqueness to generate SVMs features; (ii) a model named
Matched Peak Intensity Redistribution (MPIR) was used to handle mass modification type (fixed and variable) and cleavage
type(proper, theoretical missed and random missed) to recalculate peak intensity for each matched theoretical peptide;(iii) 17
select SVMs features are firstly calculated in crude ranking procedure to efficiently reduce the number of candidate proteins from
ten thousands to tens; (iv) robust protein prediction by a set of 491 selective and well-evaluated SVMs features; (v) dynamical
interface to easily monitor the identification pipeline; and (vi) double prediction tags and probabilities plus detailed statistics to
describe protein identification result.
Youyuan Li completed his Ph.D in Biochemical from East China University of Science and Technology (ECUST). His research interest is application
of proteomics and bioinformatics in industrial bio-processes optimization. In 2002, Li joined the faculty at School of Biotechnology, ECUST. He
teaches courses ranging from Bioinformatics to Fermentation Engineering.
Peer Reviewed Journals
Make the best use of Scientific Research and information from our 700 + peer reviewed, Open Access Journals