Fig. 2From: EPSILON-CP: using deep learning to combine information from multiple sources for protein contact predictionSimplified and aggregated depiction of the feature importance as emitted by XGBoost. The amino acid composition is attributed with the least importance, although it makes up roughly 75% of the features. The different co-evolutionary information entries correspond to different co-evolutionary methods. The feature importance depicted here is the average over a 5-fold cross-validationBack to article page