Weighted NMF for high-resolution mass spectrometry analysis

Metabolomics studies are designed to identify, measure, and interpret complex profiles of small organic molecules in different biological samples. Recent high-resolution mass spectrometers, thanks to accurate mass measurements, enable a more reliable compounds identification. The mass spectra matrices considered are nonnegative quantities exhibiting intensity-dependent noise. Among the different Nonnegative Matrix Factorization (NMF) algorithms, the weighted one seems able to handle, with a suitable weighting, this kind of noise. In this paper we compare the performance of NMF and Weigthed NMF (WNMF) in the extraction of different compounds in mass spectrometry data. Numerical experiments on simulated and real data sets illustrate the good behavior and the usefulness of this new method for high-resolution mass spectrometry.