Fig. 7: Effect of mutations on the model prediction. | Nature Communications

Fig. 7: Effect of mutations on the model prediction.

From: Peering inside the black box by learning the relevance of many-body functions in neural network potentials

Fig. 7

Panel a shows the mean 2-body relevance difference for each amino-acid pair with respect to the wild-type prediction, i.e. Rmut − RWT. The upper right half corresponds to the LEU30PHE and the lower left to the ILE4ASN mutation. A red (positive) interaction corresponds to a higher relevance in the mutated state than in the wild-type, thus a destabilized interaction. The arrows and waves correspond to regions of the protein sequence inside a β-sheet and an α-helix respectively. Panels b, c show an example structure with the mutated residue highlighted in gray as well as strongest interactions flagged by the network interpretation with colors corresponding to a. Visualizations generated with PyMOL110.

Back to article page