Peptide fragmentation prediction
Understanding peptide fragmentation, irrespective of the fragmentation method or other machine-specific settings, will enable the in-silico generation of spectra. In turn, this will allow e.g. the implementation of novel database search algorithms, the analysis of chimeric spectra or the guided generation of spectral libraries for data independent (DIA) data analysis. Making extensive use of the data generated as part of the ProteomeTools project, we are using deep learning to develop a model, which accurately predicts fragmentation spectra for any peptide of interest.
Within this project, similar techniques are used to predict other properties of peptides, such as retention time or proteotypicity. Ultimately, all models will be directly integrated into ProteomicsDB and thus support other projects like Proteomics Analytics.