Skip to main content
Fig. 3 | BMC Bioinformatics

Fig. 3

From: Predicting peptide presentation by major histocompatibility complex class I: an improved machine learning approach to the immunopeptidome

Fig. 3

Mean information by feature mirrors canonical anchor residues. The mean information (reduction in Gini impurity) across all classifiers is shown, with the inset showing only the feature subset of sparse encoding. Note that the information is higher at positions two and nine for the sparse encoding features (each a 20-dimensional encoding of each amino acid). For the nine-dimensional features of hydropathy, aromaticity, and mass, the mean information content at positions two and nine is substantially higher than at other positions. Data for this figure are provided in Additional file 5

Back to article page