iMOWSE, a scoring scheme bridging in silico and in vitro digestion in peptide mass fingerprints

Peptide mass fingerprinting is a popular protein identification method. Presence of false positives and inability to identify a protein remain a challenge. Despite much effort in tackling the problems, little attention is paid to the intrinsic problematic methodology adopted - tryptic digestion followed by identification of the experimentally available peptides. Here, we demonstrated an unexplored issue - cleavage site clusters and formulated a scoring function, iMOWSE to distinguish true positives from false positives. We first showed how the cleavage site clusters skew peptide database composition. Uneven distribution among proteins and peptides results in biased scoring. A protein falls in a group possessing a large amount of short peptides has a higher chance to score higher. This effect cannot be removed by simply limiting database search within a mass range of common experimentally obtainable peptides. Neither does artificially multiplying the factors of such short peptide matches solve the problem. Concerning the information content of an in silico digestion of a protein entry, both the number of matches of long peptides (>800 Da) and the number of such long peptides an entry can generate are taken into account. The resultant scoring function iMOWSE is robust in detecting true positives in our model database, which has a similar composition distribution to common databases msdb and nr. Moreover, it also accommodates better with situations that allow larger tolerance of deviation of query mass values from theoretical mass values; and fewer number of matches are needed to identify true positives.