ACS Fall

Despite constant advances in the QSAR field, computational approaches in general still fail to meet the expectations. In this work we attempt to re-assess the role of various in-silico tools in building an effective lead optimization strategy. From the modeling perspective, the emphasis is put on the mechanistic interpretations utilizing the most basic physicochemical characteristics of the compounds, e.g., lipophilicity, size, ionization, etc. The phenomenon of local 'anomalies'—the so-called 'activity cliffs'—is also covered, providing suggestions for the statistical methods (pairwise QSAR) suitable to resolve such issues. Automated Hansch and Free-Wilson type analysis (Auto-SAR approach) is presented as a viable solution to include target affinity data in the analysis, potentially suggesting the most promising candidates. Finally the potential of various in-silico techniques in the evaluation of other lead optimization aspects, such as synthetic feasibility or patentability prospects, is overviewed.

This work investigates the possibilities of effective third-party model utilization in predicting drug metabolism related properties. A major problem is that training sets rarely cover chemical space and experimental protocols of 'in-house' projects. A method is needed that allows any company to tailor a third-party predictive algorithm to its needs using proprietary data. We present GALAS modeling methodology that provides the possibility to expand the Applicability Domain of the models with the help of a custom database of experimental values. Use of the method is illustrated with examples of its application in predicting cytochrome P450 substrate and inhibitor specificity as well as regioselectivity. Relatively small amount (3 to 5) of similar compounds has to be added to substantially improve predictions for compounds not represented in the original training set. Similarly the models are shown to be able to utilize experimental data obtained using different protocols compared to the training set data.