What are the limitations of CLogP?

What are the limitations of CLogP?

As with any calculation algorithm, there are certain types of input that stretch the boundaries of, or complete break, the underlying model.

'Chiral' compounds:

ClogP discards all chirality specifications, because the model ClogP uses doesn't use 3D information, just the topology of the molecule. ClogP is then calculated normally, for the simplified compound.

'Charged' compounds:

CLogP will return such an estimate for charged nitrogen, phosphorous, sulfur and selenium compounds, IF NOT entered as disconnected structure (see below).

'Large' compounds:

ClogP does not attempt to calculate compound that exceed certain thresholds:

• compounds with more than 80 fragments because the inter-fragment interactions are too complex to account for.
• compounds with more than 150 heavy atoms because the model ClogP uses doesn't account for a sufficient number of intra-molecular forces.
• compounds with more than 50 flourine fragments because the compound is essentially insoluble in octanol/water.

'Disconnected' compounds:

Current practice is to select the largest carbon-containing compound from the set, and generate the ClogP based on that. This makes sense for handling salts and simple complexes, but we don't attempt to address the chemistry of mixtures at all.

Further discussion of ClogP for charged/ionic compounds:

Most ions, for which a hydrophobicity estimate is desired, are 99+% ionized in the water phase and 99+% as ion-pairs in the octanol. Therefore it is incorrect to call the ratio of concentrations between the phases as a 'partition coefficient'. As a matter of fact, it is NOT constant, but depends upon the relative concentrations. Most users want a 'relative' hydrophobicity for a congeneric series, which ClogP does provide.

CLogP will return such an estimate for charged nitrogen, phosphorous, sulfur and selenium compounds, IF NOT entered as disconnected structure. For charged nitrogen at least, this estimate will be close to the observed ratio when the concentration of the ammonium (or pyridinium) ion is very small and there is a large excess of small ion (usually chloride or bromide).

Employing it as a MODEL may be faulty if charged and neutral solutes appear together, because the charged solutes may pair up with bulky counter-ions or be attracted to proteins having charges surfaces when these are present. For example, with picrate as anion, the charges on the ion pair with tetra-alkyl ammonium solutes are effectively screened, and are, therefore, comparatively hydrophobic.