Hello,

I use PP v6.1.5 and I've noticed lately a strange behaviour, for which I'd like your insights: I used 2 methods to calculate similarities with folded ECFP_6 and Tanimoto, which, to me, were supposed to be equivalent.
The first method is using the default "Molecular Similarity" component with ECFP_6 as PredefinedSet and the parameter "OptimizeFor" is set to "Speed, 1024".
For the second method, I calculate the ECFP_6 separately and fold it to 1024 bits with the "Convert Fingerprint" component. Then, I use the Molecular Similarity using the folded ECFPs as UserSet, this time with the "OptimiseFor" parameter set to "MemoryUse" (therefore no folding).
The similarity results are supposed to be the same between the 2 methods, however there are small discrepancies.
Do you have any idea why is that?

Thanks,