Feature selection and classification of prO-TOF data based on soft information

In this paper, we introduce a feature selection and classification method for prOTOF Mass Spectrometry (MS) data profiles of diseased and healthy patients. The method is based on a special statistical measure, which quantifies the probability of the existence of peptidepeaks. A special ranking score that is based on the statistical measure is used for selecting features that can best distinguish diseased and healthy data profiles. Based on the selected features, we applied a variety of classification algorithms and the results are compared with that of a method which selects features only based on peak heights. The results show a significant improvement in classification error rate with our proposed method.