Fig. 2: Peptide MS/MS spectrum and iRT prediction.
From: In silico spectral libraries by deep learning facilitate data-independent acquisition proteomics
a A graphical illustration of the deep neural networks for peptide MS/MS spectrum and iRT prediction. b A representative spectral match of a peptide (AVLGTSNFK 2+) between its higher energy collisional dissociation (HCD) MS/MS spectrum and the corresponding in silico predicted MS/MS spectrum. c The distributions of dot products computed between predicted and experimental b/y/neutral loss peak intensities. The medians are indicated. The boxes and whiskers show the quantiles and 95% percentiles, respectively. d Pearson correlation coefficients (r) and the differences computed between predicted and experimental normalized retention time (iRT). Color gradation indicates relative density of data points. The boxes show interquartile ranges (IQR), and the whiskers show 95% percentiles; no outliers are shown. 2+: doubly charged precursors; 3+: triply charged precursors. Source data are provided as a Source Data file.
