Optimal analysis of complex protein mass spectra.

Due to physical and chemical phenomena, a simple sample can give rise to a complex mass spectrum with many more peaks than the number of molecular species present in the sample. We link peaks within and between different spectra, and come up with an advanced analysis approach to produce reliable estimates of the molecule masses and abundances. By linking peaks, we can locate multiple-charge peaks at the correct position in the spectrum, we can deconvolute complex regions with many overlapping peaks by including information from related regions with lower complexity and higher resolution, and we reduce the total number of observed peaks in a spectrum to a much smaller number of underlying molecular species. In this paper we properly model 29 952 peaks in 64 spectra, using only 39 location parameters and one shape parameter. This major reduction from many different molecules to a limited set of molecular species reduces the statistical test multiplicity for biomarker discovery and therefore we imply that the reduction should eventually increase the biomarker discovery power significantly, too.