Table 3.

Methods for protein and peptide identification

Year	Name	Neural network details	Comment	Citation
2012	Barista	special type of network or tripartite graph where layers represent proteins, peptides, and spectra	protein inference through integration of protein and peptide identification	Spivak et al. (2012)
2017	DeepPep	CNN, torch7 framework	predicts peptide probability from binarized protein sequence, protein scored based on change in peptide prediction without each protein	Kim et al. (2017)
2017	DeepNovo	LSTM/CNN hybrid network built with TensorFlow	application to DDA data. Iteratively predicts one amino acid at each step. Up to 64% better than previous algorithms	Tran et al. (2017)
2018	DeepMatch	bidirectional LSTM, weak supervision	spectral prediction integrated with peptide identification	Schoenholz et al. (2018)
2019	DeepNovo	LSTM/CNN hybrid network built with TensorFlow	adapted to DIA data by incorporating the retention time dimension	Tran et al. (2018)
2020	DIA-NN	ensemble of dense, feedforward classifiers. Implemented with Cranium DNN	operates with or without a user-supplied spectral library	Demichev et al. (2020)
2020	DeepRescore	uses AutoRT and pDeep2 models	generates new scores derived from comparing observed peptide properties to deep learning-predicted properties. Those scores are input to Percolator	Wen et al. (2020b)