![\"\"](\"/assets/images/depict2060.png\")
The readers of Antony’s blog\nknow enough about the problem. And many in the QSAR community know it too (and\nmany other do not). Chemical structure data is noisy. I haven’t recently created\na new local data set for analysis, so I have not taken time to blog about it\nmuch, but the ambiguity in chemical databases is enormous. Just yesterday, Antony\nand I had a good\ndiscussion\nabout tautomers and in particular how things are linked together.
\n\nIf we are in the field of property prediction, knowing what tautomer to calculate\ndescriptors for is crucial. Not that we actually have easy access to experimental\ndata showing what the important tautomer is for our end-point (predicted property),\nbut at least we can track what tautomer we modeled with. Has everyone ever asked\nyou to add units to experimental values? Like “the temperarature was 279 degrees;\nCelsius or Kelvin??” Well, this is the exact same thing. If your QSAR model\ntraining report does not include that information, you are doing it wrong. (/me\nducks)
\n\nSo, why does it in fact matter? It matter simply because calculated properties\nare different. Backing up to the ChemSpider example in my question about InChIs\nwith the fixed-hydrogen layer\nI noted that (like in many other databases) the synonyms seems to include IUPAC\nnames for at least two tautomers. However, while the ChemSpider is, in fact,\nfor the tautomer-independent structure (using the InChI mobile hydrogen layer;\nand keep in mind that the InChI uses only a limited amount of heuristic rules\nfor identifying tautomers, making it not detect all 40 tautomers of warfarin),\nthe 2D diagram, the 3D model, and the calculated properties reflect only one\ntautomer.
\n\nAnd calculated properties are exactly the input in QSAR’s statistical modeling.\nIt is interesting to realize that the differences in calculated molecular descriptors\ncan vary both minimally, or not at all, as drastically. Very drastically, in\nfact. The recent paper by Porter (doi:10.1007/s10822-010-9335-7)\nshows the 40 warfarin tautomers, and discusses a few properties, such as the\npKa. The experimental pKa of warfarin is around 5. Now, the paper reports calculated\npKa values for a variety of software products (AMBIT\nis unfortunately missing). First of all, it shows that the various tools differ,\nwhich is to be expected. But that variance is neglectable when compared by the\neffect of picking the wrong tautomer. I was impressed by the range of predicted\nvalues for the various tautomers. I ranged from about 5 to 12, throughout all\ntools. That means warfarin is predicted to be mildly acidic (some tools predict\npKa’s down to 2.5) to very basic! No way your statistical modeling will understand\nthat!
\n\nAnd this is why Open Data is so important in chemistry. So, the next time Joe\n(Organic) Chemist bitches about computers and cheminformatics, tell him it is\nhis own fault: he should have released his data out in the Open.
\n\nAnyway. Tautomerism was a curation issue in the first(!!!) entry I was curating.\nThe sixth had the more well-known problem, I think. I may be blind, but I would\nsay this drug has\na stereocenter:
\n\n
But none of the databases I checked so far (including ChemSpider) defines the\nstereochemistry! I thought we settled that some decades ago? Stereochemistry\nof drugs matter. What is going on here? I guess I have to browse some primary\nliterature and access some experimental data today then. If I can afford it.
\n\n