On the preprocessing of mass spectrometry proteomics data