The feature selection approach for evaluation of potential rheumatoid arthritis markers using MALDI-TOF datasets

Abstract

Background: The selection of the most representative mass profiles, in rheumatoid arthritis (RA) serum samples was developed. This allows for selection and identification of potential biomarkers in RA serum samples.

Methods: The RA and controls samples were analyzed using MALDI-TOF. Two different protein elution procedures utilizing ZipTips (E1 and E2) were examined. The statistical evaluation of data was performed using different feature selection (FS) methods in combination with different classifiers, while identification of selected masses was performed using MALDI-TOF-TOF.

Results: Utilization of proposed statistical strategy allowed for the selection of different masses according to FS method and elution procedure. Obtained masses were further subjected for targeted identification. The panel of proteins were identified as potential markers. The role of these proteins was discussed in relation to pathomechanism of RA.

Conclusion: Application of advanced biostatistical analysis of obtained MALDI-TOF datasets, resulted with targeted selection of potential RA biomarkers. Five proteins were identified due the E1 procedure, and six proteins were identified due the E2 procedure, respectively. The panel of identified proteins suggest that presented statistical methodology and proteomic strategy was correct and gave valid results. Obtained results may contribute to development of clinically useful multicomponent diagnostic tool.

Keywords: Biomarkers discovery; Feature selection statistical strategy; MALDI-TOF; Peptide profiling; Rheumatoid arthritis.