Extended Data Fig. 1: Evaluation of CryoBench dataset results for the IgG-RL and MDSpike datasets.
From: Merging conformational landscapes in a single consensus space with FlexConsensus algorithm
Panel a) shows the evaluation of the correlation between the FlexConsensus latent space vectors obtained for the IgG-RL dataset from CryoBench11 and two internal variables (d and θ) designed to characterize simulated disordered motions (the Mantel test used to evaluate the significance of this correlation was two-sided, yielding a p-value of 0.002). Panel b) shows the comparison of the location of the collective variables measured from the ground truth energy landscape of the MDSpike dataset from CryoBench11 in its original energy landscape and FlexConsensus space. Three landscapes were input to the consensus: the ground truth energy landscape, the conformational landscape of HetSIREN, and the conformational landscape of CryoDRGN. The results highlight the correlation of the original collective variables (CV1 and CV2) with the consensus space, which are almost aligned with the consensus space’s principal components (PC).
