Table 1.

Methods for fragment ion intensity prediction

Year	Name	Neural network details	Comments	Citations
2005	PeptideART	feedforward network	engineered peptide feature inputs, outputs of fragment probabilities	Arnold et al. (2005), Li et al. (2011)
2017	pDeep	bidirectional LSTM, multi-output regression; Keras v1.2.1, TensorFlow 0.12.1	limited to peptides of up to 20 amino acids	Zhou et al. (2017)
2018	DeepMatch	bidirectional LSTM, weak supervision	direct integration with peptide spectrum matching algorithm outperforms COMET	Schoenholz et al. (2018)
2018a	Prosit (latin for “of benefit”)	encoder: bidirectional GRU with dropout and attention, parallel encoding of precursor charge and collision energy; decoder: bidirectional GRU with dropout and time-distributed dense; multi-output regression Keras 2.1.1 and TensorFlow 1.4.0	over half a million training peptides and 21 million MS/MS spectra at multiple collision energies, predicts MS/MS spectra and retention time, integration with database search to decrease FDR, integration with Skyline (MacLean et al., 2010), web tool https://www.proteomicsdb.org/prosit/	Gessulat et al. (2019)
2019a	DeepMass	encoder: three bidirectional LSTM with 385 units each; decoder: four fully connected dense layers 768 units each; multi-output regression TensorFlow v.1.7.0	predicted fragmentation with accuracy similar to repeated measure of the same peptide's fragmentation. Predicted spectra used for DIA data analysis nearly equivalent to spectral libraries	Tiwary et al. (2019)
2019	pDeep2	bidirectional LSTM, multi-output regression	original pDeep model adapted to predict spectra of modified peptides using transfer learning	Zeng et al. (2019)
2019a	N/A	encoder: bidirectional LSTM with dropout; iRT model, two dense layers, tanh, single output regression. Charge state distribution model, two dense layers, softmax activation, multi-output regression length 5 for charge 1–5. Spectral prediction model, a time-distributed dense layer with sigmoid activation function, multi-output regression; Keras	predicts retention time, precursor charge state distribution, and fragment ion spectra	Guan et al. (2019)
2019	MS2CNN	basic CNN architecture, engineered peptide features as input with a CNN kernel size of 4	better than pDeep for prediction of spectra from +3 charge state peptide precursors	Lin et al., 2019
2020a	DeepDIA	hybrid CNN and bidirectional LSTM, CNN first extracts features from pairs of amino acids, then LSTM, then dense layer. Multi-output regression of the b/year ions, including water/ammonia losses. Keras 2.2.4 and TensorFlow 1.11	predicts MS/MS spectra and indexed retention time (iRT). Slightly more protein identifications from DIA analysis of Hela proteome than libraries from DDA or Prosit	Yang et al. (2020)
2020	N/A	sequence-to-sequence CNN	full-spectrum prediction, not only fragment ions	Liu et al. (2020)

Abbreviation are as follows: FDR, false discovery rate; N/A, not applicable.

^aIndicates methods that predict other factors apart from fragment ion spectra.