An improved algorithm for peak detection in mass spectra based on continuous wavelet transform

Peak detection is an important preprocessing step in the analysis of mass spectrometry data. Currently, most peak detection methods have limited identification ability when there are overlapping or low amplitude peaks. In this paper, we present an improved algorithm that combines the continuous wavelet transform (CWT) with the crazy climber algorithm. Particles move on the CWT coefficient matrix according to certain rules, and gradually gather at local maximal points. Peaks are identified by drawing ridges based on the weighted occupation measure of the grid points. The proposed method is evaluated using simulated noisy spectra and real spectra of benzene in nitrogen. The results show that the proposed approach is better at identifying overlapping peaks than existing methods. Receiver operating characteristic curves show that the proposed method can detect more true peaks while maintaining a low false discovery rate. (C) 2016 Elsevier B.V. All rights reserved.