Compound PCLs were also assessed for structural similarity. The 2D structural similarity of all pairwise combinations of compounds within each PCL was measured using Tanimoto coefficient calculated from binary fingerprints, which were obtained from SMILES strings representing structures of the compounds in PCLs. SMILES strings were converted to binary fingerprints using the Open Babel implementation of the Daylight fingerprint standard (O'Boyle et al., 2011). We found that the vast majority of PCLs were structurally diverse. All but one PCL had a median pairwise Tanimoto below 0.8. Detailed information on all PCLs is available in Supplementary Table S7.
