Fig. 5: Prosit rescoring of 25 patient melanoma samples. | Nature Communications

Fig. 5: Prosit rescoring of 25 patient melanoma samples.

From: Deep learning boosts sensitivity of mass spectrometry-based immunopeptidomics

Fig. 5

a Vennbars showing the number of peptides lost (red), shared (blue) and gained (green) by the MaxQuant rescoring pipeline compared to the results published by B.-Sternberg and Bräunlein et al.11 for each patient. b Venn diagram comparing identified mutated neoepitopes by MaxQuant only (red) and by Prosit-based rescoring of MaxQuant only (dark green) and shared identifications (blue). Peptides selected for synthesis and further testing are depicted in light green. c Results of ELIspot immunogenicity assays probing synthetic mutated neoepitopes against Mel15 derived PBMCs. Readout was the number of spot forming cells probing interferon-gamma secretion. All batches and replicates were normalized to the respective positive controls. Between 3 and 6 replicates are plotted for each peptide. A KIF2C wildtype peptide, as well as unstimulated cells, served as controls. Assay data of replicates of the mutated KIF2C peptide RLFLGLAIK are highlighted in green. Mutations are shown as bold underlined amino acids. Raw and analysis data are available from the PRIDE repository with identifiers PXD021398, PXD004894, and Supplementary Data 7.

Back to article page