Using neural networks to identify more proteins in high-throughput proteomics

Protein identification using mass spectrometry is a critical step in many areas of the life sciences, and in proteomics in particular. To confirm the presence of a protein in a sample, at least one of the constituent peptides from that protein must be matched to a theoretical peptide sequence. The prediction of a fragmentation spectrum from a theoretical sequence so that it can be compared to an observed spectrum is the key challenge of protein identification algorithms. We present a study using artificial neural networks to learn properties of fragmentation spectra so that more peptides and therefore proteins can be identified in high-throughput proteomics.