Fig. 3: Autoencoders and DiffNets can both compress protein structures and then reconstruct them. | Nature Communications

Fig. 3: Autoencoders and DiffNets can both compress protein structures and then reconstruct them.

From: Deep learning the structural determinants of protein biochemical properties by comparing structural ensembles with DiffNets

Fig. 3

a Reconstruction error plots showing the root-mean-square deviation (RMSD) between a protein structure from simulation and the corresponding protein structure generated by unsupervised autoencoders (yellow) or DiffNets (blue). One of every ten structures from wild-type and M182T simulation data was used (n = 65,210) and the standard deviation is shown with error bars. b Structure representing the difference between a structure generated by the DiffNet (blue) vs. the actual conformation from simulation (gray) when training with 3 latent variables.

Back to article page