Fig. 5
Clustering of hSGLT1 actives. Active hSGLT1 compounds in the training set clustered into ten chemical clusters (Tanimoto, FCFP6). Molecular structure and affinity (pKi for public data and  % of (negative) control for in-house data) for representative cluster compounds are shown. In-house compounds with activity < 70% of (negative) control and public compounds with pChEMBL ≥ 6.5 were used in clustering. a t-SNE plot of the chemical clusters. b The molecular weight and ALogP distribution of compounds in the chemical clusters