Analysis of MALDI-TOF Serum Profiles for Biomarker Selection and Sample Classification

Mass spectrometric profiles of peptides and proteins obtained by current technologies are characterized by complex spectra, high dimensionality, and substantial noise. These characteristics generate challenges in discovery of proteins and protein-profiles that distinguish disease states, e.g. cancer patients from healthy individuals. A challenging aspect of biomarker discovery in serum is the interference of abundant proteins with identification of disease-related proteins and peptides. We present data processing methods and computational intelligence that combines support vector machines (SVM) with particle swarm optimization (PSO) for biomarker selection from MALDI-TOF spectra of enriched serum. SVM classifiers were built for various combinations of m/z windows guided by the PSO algorithm. The method identified mass points that achieved high classification accuracy in distinguishing cancer patients from non-cancer controls. Based on their frequency of occurrence in multiple runs, six m/z windows were selected as candidate biomarkers. These biomarkers yielded 100% sensitivity and 91% specificity in distinguishing liver cancer patients from healthy individuals in an independent dataset.