Fig. 7.

Average ranks of 2048-D ECFP4/6, MHFP4/6 and SECFP4/6 across 88 benchmark targets. The SMILES-based circular substructure hashing scheme applied by MHFP was folded using ECFP’s module $n$ method. This fingerprint variant was denoted SECFP. While SECFP4/6 were outperformed by MHFP4/6 respectively, SECFP6 performed significantly better than both ECFP4/6. These results suggest that SECFP6 can be readily used as a drop-in replacement for ECFP with beneficial results. By performing significantly worse compared to MHFP6, acting as a control, SECFP6 further validates the minhashing approach as compared to folding