Dear Noel, I recently started to look at descrptor values from various toolkits, without initially being aware of a two years old blog post of yours: http://baoilleach.blogspot.com/2008/04/cheminformatics-toolkit-face-off-logp.html
I used paracetamol and another test compound (mol1) in order to make comparisons. The result is: Paracetamol =========== * Pybel TPSA: 49.330000 LogP: 1.423600 * RDKit TPSA: 49.330000 MolLogP: 1.350600 * CDK TPSA: 49.33 ALOGP: -1.0852000000000002 ALOGP2: 1.1776590400000004 XLOGP: 1.079 Mol1 ==== * Pybel TPSA: 117.560000 LogP: 4.719300 * RDKit TPSA: 83.880000 MolLogP: 4.629500 * CDK TPSA: 117.56 ALOGP: 1.4521000000000013 ALOGP2: 2.108594410000004 XLOGP: 0.7899999999999999 Mol1 is described by this smiles chain: CC(=O)c1ccc(/N=C2\SCCCN2C(=O)c2ccc(CS(C)(=O)=O)cc2)cc1 Even TPSA is not consistent across toolkits. The diversity of LogP implementations makes difficult to use published works that rely on LogP values for the molecule property prediction, especially when descriptors are calculated by some not freely available software. More precisely, I refer to a recently published work Journal of Molecular Graphics and Modelling 28 (2010) 899?903 that involves LogP(o/w) as calculated by MOE, which is said to be different from SLogP when calculated by MOE and by RDKit. What a jungle for a newcomer in the field! As the author of Cinfony, I would like to have your opinion about the origin of different values I obtained for TPSA on compound "mol1". Best regards, Jean-Marc Nuzillard ------------------------------------------------------------------------------ _______________________________________________ OpenBabel-discuss mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/openbabel-discuss
