Abstract
Topological methods for molecule volume prediction based on local fragments centered on all carbons of aliphatic alkanes (C6-C11, C12) are proposed and evaluated. The LOGIC method dealing with local to global information construction is based on local environment (FRELB) partition of information, both structural and physical, and the use of these FRELs to correlate the molal volumes V. Different topological vectorial sets of various depths, dealing with atom and bond FRELs, are used to optimize the contributions V D of a set of prime FRELs constituting the reference set structured in a hyperstructure linking 70 FRELB. The accuracy of determining fragment volumic contributions is traced by the decreasing number of hits resulting in the modeling of the C6-C11 populations. The training set of the V values is tested on the 355 alkanes of the C12 population. This extrapolation leads to excellent statistical results of the LOGIC method, which is a “parent structure free” similarity correlation although its application here is “aliphatic alkane” oriented. The set of V D values is obtained through the combined use of several topological sets and the resolution of FREL intersections to control redundant structural information. In practice its application is simple and strictly additive.