Modeling alcohol metabolism with the DARC/CALPHI system

We present our general system for QSAR search, CALPHI (Computer-Aided Law by Hyperstructure Investigation) set up in the context of the DARC structural language. We use it to construct global, fragmentary, and topological models of the capacity of alcohols to undergo glucuronidation. The DARC/PELCO model, more precisely and more significantly, explains 98% of the total variance with only three parameters, while treating the whole set of primary, secondary, and tertiary alcohols, whereas the best previously reported treatment restricted to primary alcohols, explains only 90% of the variance with two parameters. It provides an explicit and more precise interpretation of alcohol metabolism. The PELCO methodology is extended to evaluate the prediction reliability of both global and fragmentary models. PELCO leads to more predictions when comparison is made at the same level of reliability.