Fig. 3From: Deep clustering of small molecules at large-scale via variational autoencoder embedding and K-meansEmbedding-specific silhouette scores under different number of molecule clusters using the global features. Results are based on the 193 global features and their 64, 32, and 16 embeddings from the VAE and AE algorithms, respectively. It should be noted that 7 raw global features with low variation were filtered out (see “Materials and methods” section)Back to article page