Fig. 7From: Deep clustering of small molecules at large-scale via variational autoencoder embedding and K-meansTanimoto similarity matrix between each pair of the examples, including the reference compounds (Mol0) and three test compounds (Mol1, Mol2, Mol3). The binary ECFP (radius = 1, bit = 2048) were used for the similarity calculation. The orange rectangle circles the samples belonging to the same clusterBack to article page