Calculation of Abraham descriptors from experimental data from seven HPLC systems; evaluation of five different methods of calculation

摘要

Most of the recent work in ‘property-based design’ to convertna lead or candidate molecule into a successful drug on thenmarket, has highlighted the importance of molecular size andnhydrogen bonding capacity as a way to understand ‘druglikeness’.n1 Various methods have been employed to predict biologicalnproperties that are difficult or costly to measure in thenfirst stages of drug discovery, but in many multi-componentnapproaches the descriptors used are poorly defined and/or arendifficult to relate to specific structural elements of the propertiesnof functional groups. The need to obtain well definednexperimental molecular descriptors for drug compounds by fastnand efficient processes to suit the industry remains of primarynimportance.nThe Abraham method is one such method.2–5 It is based onnthe solvation equation (1), which correlates solute propertiesn(SP), such as partitioning,6,7 solubility,8 blood-brain distributionn9 and human intestinal absorption,10 with a standard setnof five molecular descriptors. Although descriptors E (Excessnmolar refraction) and V (McGowans volume) can be obtainednfrom structure, the model’s use is restricted due to the difficultynof measuring the molecular descriptors; S (dipolaritypolarisability), A (hydrogen bond acidity) and B (hydrogenbondnbasicity). The solute descriptors represent the soluteninfluence on various solute–solvent phase interactions. Hencenthe regression coefficients c, e, s, a, b and v correspond to thencomplementary effect of the phases on these interactions. Then† Electronic supplementary information (ESI) available: Tables S1 tonS5. See http://www.rsc.org/suppdata/p2/b2/b206927jcoefficients can then be regarded as system constants whichncharacterize the phase and contain chemical information aboutnthe phase in question.nThe experimental method often used to determine S, A and Bnis through the use of water/solvent partition (log P) measurements.nThe method has been recently reviewed,11 and the errorsninvolved have been evaluated. Although the results have provennto be satisfactory, the procedure itself remains relatively slownfor industrial purposes due to lengthy sample preparation. Innconcurrent studies, Valko12–14 and co-workers have used the fastngradient elution RP-HPLC in order to obtain Abrahamndescriptors in a much more rapid procedure. Valko andnco-workers have set up their method by choosing the mostnorthogonal HPLC systems by non-linear mapping. As anstraight forward extension of our earlier work11 we apply thenmathematical methods we have described to the systems chosennby Valko and co-workers.12–14nThe descriptors for the 80 solutes used are shown in Table 1.nThese descriptors have been calculated from a variety of equationsnwhich included equations for chromatographic data, asnwell as for a very large number of log P values. We will refer tonthe descriptors obtained in this way as ‘Table 1’ descriptors. ThenRP-HPLC data used are in the form of CHI (chromatographicnhydrophobicity index) values; these are derived from thenexperimental log k values using the standard procedure asndescribed by Valko and co-workers.12–14 In Table 2 we tabulatenthe CHI values for the 80 solutes on the seven systems. Theseven systems used in this study have been chosen to cover thenwidest range of the coefficients for A, B and S in the solvationnequation. Most of the chosen columns are stable at relativelynhigh pH (10.5) that is important from a practical point ofnview, as the compound’s retention should be measured at anpH where it is not ionised. Thus, the retention times of basicncompounds should be measured with high pH mobile phases.nWe have chosen short columns and fast generic gradients withn5 minutes cycle time. It means that gradient retention timesn(CHI values) for one compound can be obtained in 35 minutesnunder 7 different HPLC conditions. The systems themselves arenlisted in Table S2 (Supplementary Information †).