The new algorithm allows gains in confidence and reproducibility and will be released as part of Thermo Fisher’s newest Thermo Scientific Proteome Discoverer 2.5 software release. Users can now access deep-learning-based prediction of tandem mass spectra, allowing for the formation of entire spectral libraries on demand and facilitating the identification of peptides with up to 10 times higher confidence and the extraction of more identifications from proteomics datasets via intensity-based rescoring. In combination with Thermo Scientific Orbitrap technology, the new algorithm enables emerging applications, such as immunopeptidomics and metaproteomics, for which traditional database search and statistical approaches are often ineffective.
"Increasing the confidence of protein and peptide identifications is a growing need, given that a false discovery rate of even 1% means that 1,000 out of every 100,000 peptides might be incorrectly assigned," said Mark Sanders, director of life science mass spectrometry software, Thermo Fisher Scientific. "Applying deep learning tools enables data-independent analysis of proteomics samples with higher confidence and reproducibility, and, when used with Orbitrap technology, reduces the false discovery rate 10-fold, to merely 100 out of every 100,000 peptides."
Martin Frejno, chief executive officer, MSAID GmbH, said, "At MSAID, we reinvent the way proteomic data is acquired and analyzed by using state-of-the-art deep learning. Through our collaboration with Thermo Fisher Scientific, we can bring this technological revolution to laboratories around the world and empower the scientific community to gain exceptional insight into new and existing data."
Thermo Fisher Scientific will showcase outcomes of the collaboration and its newest products and software solutions in a company-hosted virtual event, vLC-MS.com (May 26-28, 2020) and at the American Society for Mass Spectrometry (ASMS) Reboot Program (June 1-12, 2020).