I've calculated logP in Marvin6.0.3 (using default settings - weighted method; electrolyte concentration - 0.1; ticked Consider tautomerization/resonance) for two very similar type II kinase inhibitors. The only difference is in the linker which is -O- atom in one cpd and -NH- in the other compound. The difference in the logP is substiantial: logP for -O- compound is 5.05 and for -NH- cpd is 6.29. Now, the atom contributions (atom increments) are identical for both cpds and NH has even lower contribution than -O- (which is reasonable, you'd expect slightly lower logP for -NH- cpd). But the value of structural increment for -NH- compound is 1.95 (0.65 for -O- compound), and this is responsible for the large difference in final logP values. What is this structural increment? I searched chemaxon webpage and went through the original Viswanadhan JCIM paper and couldn't find explanation. Is this a correction for electron delocalization? Linkers link two aromatic rings.
I also caluclated logP values with Qikprop from Schrodinger and there are serious differences, see table below:
linker marvin qikprop
-O- 5.05 3.05
-NH- 6.29 2.80
The differences between -O- and -NH- cpds in qikprop are much smaller, as you'd expect. Why there are so large differences between -O- and -NH- in marvin and what is the structural increment responsible for? I don't ask about differences in logP values between marvin and qikprop as I assume different methods are used, although it would be nice to know which estimate is closer to the real value...