An approach to assessing peptide mass spectral quality without prior information

This paper proposes an approach to assessing the quality of tandem mass spectra without any prior information. The proposed approach includes: filtering noises from the experimental mass spectra and extracting the peaks; mapping each spectrum into a feature vector which describes the quality of spectra; classifying spectra into clusters by using the mean-shift clustering; learning a classifier using the two clusters with the extreme means; assessing all spectra by using the trained classifier. Computational experiments illustrate that the proposed approach can eliminate majority of poor quality spectra while losing very minority of high quality spectra.