Differentiate two similar fingerprints

User 21740a6b22

11-01-2016 17:39:32


I want to differentiate between two similar molecules (SMILES pasted below) using Tanimoto and Euclidean methods. I have tried various fingerprinting parameters. However, both dissimilarity scores always amount to 0, meaning the two molecules cannot be differentiated computationally. Can the technical support please provide some advice? Thank you in advance!

Here are the settings I have tried:

ChemAxon version: 5.9

Fingerprint length: 128, 256, 512, and 1024

Maximum pattern length: 7 and 8

Bits to be set for patterns: 2 and 3

Two molecules to be distinguished:

molecule1: CCC(CC)C(=O)O[C@H]1C[C@@H](C)C=C2C=C[C@H](C)[C@H](CC[C@@H]3C[C@@H](O)CC(=O)O3)[C@@H]12

molecule2: CCC(C)(C)C(=O)O[C@H]1C[C@@H](C)C=C2C=C[C@H](C)[C@H](CC[C@@H]3C[C@@H](O)CC(=O)O3)[C@@H]12

The fingerprints were generated using the code (new MolHandler(molecule1)).generateFingerprintInInts(param1, param2, param3)

ChemAxon 8b644e6bf4

13-01-2016 12:59:35


The default chemical fingerprint represents linear paths. It can no differentiate structures containing the same set of such features. (The current implementation can not consider occurrence counts.)

The given structures however are considered different by using the ECFP fingerprint family, even with its default parametrization. At a first glance it seems that our most recent Maccs166 fingerprint implementation also differentriate these two structures, however its discrimination power can be considered generally weaker than the two other fingerprints.

Could you tell us more about your use case?