A Taxicab geometry quantification system to evaluate the performance of in silico methods: a case study on adenosine...

Among still comparatively few G protein-coupled receptors, the adenosine A_2A receptor has been co-crystallized with several ligands, agonists as well as antagonists. It can thus serve as a template with a well-described orthosteric ligand binding region for adenosine receptors. As not all subtypes have been crystallized yet, and in order to investigate the usability of homology models in this context, multiple adenosine A₁ receptor (A₁AR) homology models had been previously obtained and a library of lead-like compounds had been docked. As a result, a number of potent and one selective ligand toward the intended target have been identified. However, in in vitro experimental verification studies, many ligands also bound to the A_2AAR and the A₃AR subtypes. In this work we asked the question whether a classification of the ligands according to their selectivity was possible based on docking scores. Therefore, we built an A₃AR homology model and docked all previously found ligands to all three receptor subtypes. As a metric, we employed an in vitro/in silico selectivity ranking system based on taxicab geometry and obtained a classification model with reasonable separation. In the next step, the method was validated with an external library of, selective ligands with similarly good performance. This classification system might also be useful in further screens.