Development of a peak extraction method using the high-resolution matrix-assisted laser desorption/ionization time-of-flight mass spectrometry and machine learning techniques: Analysis of peak shapes.

Abstract

Combining matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOFMS) and Kendrick mass defect (KMD) analysis is a powerful tool for visualizing polymers in complex mass spectra. The identification of minor polymers by KMD analysis requires reduction of the broad noise peaks often observed in the low-mass region. A machine-learning model was created using pix2pixHD. It converts an original mass spectrum into a pseudo-mass spectrum that contains only the original peaks at m/z positions that the model judges as sharp single-component peaks. It reduces noise by selecting only the m/z and intensity values from the original spectrum's peak list that correspond to peaks in the pseudo-mass spectrum. A machine-learning model was applied to a low-concentration polymer mass spectrum observed at m/z <2000. Extracting single-component peaks from the mass spectrum made the minor polymer series appear clearly in the KMD plot. The technique facilitated mass spectrometric imaging of the ultraviolet degradation of polyethylene terephthalate by plotting the polymers' spatial distributions. It could also distinguish between polymer series (before and after degradation) to identify their separate spatial distributions. A machine-learning method for peak extraction from high-resolution MALDI-TOFMS was developed. Single-component peaks of the mass spectrum were distinguished from noise peaks by their peak shapes. Combining with KMD analysis facilitated the identification of minor polymer series in complex mass spectra.