Skip to main content
. 2021 Feb 19;12:1185. doi: 10.1038/s41467-021-21352-8

Fig. 4. Deep learning peptide CCS values.

Fig. 4

a Architecture of the neural network. Bi-directional long short-term model (LSTM): (i) amino acid sequence input, (ii) vectorization of amino acid information for processing, (iii) bi-directional LSTM layers, (iv) reduction to fixed length peptide feature vector by concatenating the last output neurons of both directional LSTMs, and (v) CCS prediction. b Relative deviation of predicted CCS values from an independent experimental validation dataset of synthetic peptides from the ProteomeTools project. c Correlation of predicted versus experimental CCS values (n = 155,004). d Dependence of the median relative error on training dataset size. e Same for Pearson correlation coefficient. Source Data are provided as a Source Data file.