Figure 9.

Results for alanine dipeptide in vacuum at 300 K. Clustering of the PT simulation data for the different embeddings. The results show how the embeddings map the metastable states. The data points are colored accordingly to their cluster. The first panel shows the metastable state clusters in the (Φ, Ψ) space. The second panel shows the results for the PT embedding. The third and fourth panels show the results for a representative case of a WT-MetaD embedding (γ = 5), obtained with and without incorporating weights into the training via a reweighted feature probability distribution (see eq 9), respectively. For the details about clustering,⁹⁶ see Section S5 in theSupporting Information. The units for the MRSE embeddings are arbitrary and only shown as a visual guide.