In tandem mass spectrometry-based proteomics, proteins are digested into peptides by specific protease(s), but generally only a fraction of peptides can be detected. To characterize detectable proteotypic peptides, we have developed a series of methods to predict peptide digestibility and detectability. Here, we propose a bidirectional long short-term memory (BiLSTM)-based algorithm, named DeepDetect, for the prediction of peptide detectability enhanced by peptide digestibility. Compared with existing algorithms, DeepDetect is featured by its improved prediction accuracy for a wide range of commonly used proteases, covering trypsin, ArgC, chymotrypsin, GluC, LysC, AspN, LysN, and LysargiNase. On 11 test data sets from E. coli, yeast, mouse, and human samples, DeepDetect achieved higher prediction accuracies than PepFormer, a state-of-the-art deep-learning-based peptide detectability prediction algorithm. The results further demonstrated that peptide digestibility can substantially enhance the performance of peptide detectability predictors. As an application, DeepDetect was used to reduce the in silico predicted spectral libraries in data-independent acquisition mass spectrometry data analysis. Experiments using DIA-NN software showed that DeepDetect can significantly accelerate the library search without loss of peptide and protein identification sensitivity.

• PMID: 36908083
• DOI full text
• PubMed
• PubTator

Reference Type

Journal Article | Research Support, Non-U.S. Gov't

Authors

Yang J, Cheng Z, Gong F, Fu Y

Primary Lit For

Additional Lit For

Review For