Extracting salient information from mass spectra of prostate cancer

This paper presents an application of multilevel wavelet analysis for high dimensional mass spectrometry data. Low frequency (approximation) coefficients, which contain major information contents of the mass spectra data, are extracted. Approximation of the spectra is reconstructed based on orthogonal wavelet approximation coefficients for locating the key m/z values of the mass spectra. Genetic algorithm is then used to select best features from the reconstructed approximation. Good performance is achieved and the corresponding significant m/z values of mass spectra are identified based on selected features.