Protein identification from mass spectrometry is typically performed in a so-called bottom-up fashion. In this approach observed peptide mass spectra are matched to computationally predicted mass spectra. Scoring is typically based on the mass difference between observed and predicted peaks and on the intensity of the explained experimental spectrum. Only recently has there been progress in accurate theoretical intensity prediction, using deep neural networks.

The goal is to find out how useful these predicted intensities are in scoring peptide candidates.

(Adapted from Tjeerd's description for requesting access to a GPU)