Investigating experimental vs. Predicted pK_a values for PET radiotracer

The prediction of central nervous system (CNS) active pharmaceuticals and radiopharmaceuticals has experienced a boost by the introduction of computational approaches, like blood-brain barrier (BBB) score or CNS multiparameter optimization values. These rely heavily on calculated pK_a values and other physicochemical parameters. Despite the inclusion of various physicochemical parameters in online data banks, pK_a values are often missing and published experimental pK_a values are limited especially for radiopharmaceuticals. This comparative study investigated the discrepancies between predicted and experimental pK_a values and their impact on CNS activity prediction scores. The pK_a values of 46 substances, including therapeutic drugs and PET imaging radiopharmaceuticals, were measured by means of potentiometry and spectrophotometry. Experimentally obtained pK_a values were compared with in silico predictions (Chemicalize/Marvin). The results demonstrate a considerable discrepancy between experimental and in silico values, with linear regression analysis showing intermediate correlation (R²_(Marvin) = 0.88, R²_{(Chemicalize)} = 0.82). This indicates that if one requires an accurate pK_a value, it is essential to experimentally assess it. This underscores the importance of experimentally determining pK_a values for accurate drug design and optimization. The study's data provide a valuable library of reliable experimental pK_a values for therapeutic drugs and radiopharmaceuticals, aiding researchers in the field.