Descriptors | ECFP | FCFP | logP | MW | RTB | HBA | HBD | HAC |
|---|
model 1 | X | | | | | | | |
model 2 | | X | | | | | | |
model 3 | X | | X | X | X | X | X | |
model 4 | X | | X | | | X | X | X |
model 5 | | X | X | X | X | X | X | |
model 6 | | X | X | | | X | X | X |
- ECFP unhashed RDKit Morgan fingerprint with radius of 3, useFeatures parameter set to False and useCounts set to True, FCFP unhashed RDKit Morgan fingerprint with radius of 3, useFeatures parameter set to True and useCounts set to True, logP molecular lipophilicity [20], MW molecular weight, RTB number of rotatable bounds, HBA number of hydrogen bound acceptor, HBD number of hydrogen bound donor, HAC Heavy atom count
- The non-binary descriptors (logP, MW, RTB, HBA, HBD and HAC) were discretised into 10 bins. Unhashed ECFP and FCFP fingerprints were used, so the numbers of bit descriptors varied according to size and chemical diversity of each individual dataset. The resulting data were stored in sparse matrix objects for more efficient processing
