On the development and application of AIBL-pKa, a pKa predictor based on equilibrium bond lengths of a single protonation state

Access to files

•   FULL-TEXT.PDF (pdf)

Abstract

Development of a new drug or agrochemical product is a multifaceted task, and it often requires many years of research and millions of pounds to get a single compound to market. During the discovery process, thousands of compounds are screened for their pharmacokinetic properties, bioavailability and toxicity. As the ionization state of a compound at specific pH can influence such properties, knowledge of its aqueous acid dissociation constant(s) (pKa) provides a vital tool in understanding and predicting efficacy and mechanism of action. In silico methods of pKa prediction are now a vital part of modern drug and agrochemical discovery, as in addition to saving time and materials, they allow for virtual screening of millions of compounds to take place, very early in the discovery process. The AIBL-pKa approach (Ab Initio Bond Lengths-pKa) is a pKa prediction method, which works on the basis that for a series of electronic congeners, certain equilibrium bond lengths have a linear relationship with their aqueous pKa values, even when modelled in the gas-phase. Whilst many pKa prediction methods exist, each having their own caveats and advantages, there are some types of compound for which predictions remain intrinsically challenging. Problematic compounds include those that exhibit tautomerism, compounds which have 50+ atoms and high conformational felixibility, and compounds containing multiple sites of ionization. The research described here shows that AIBL-pKa can provide solutions to these more complex challenges, with Mean Absolute Error values for external test sets typically below 0.35 log units. Furthermore, our use of quantum chemically derived 3D structures means that hydrogen bonding and steric effects on pKa are implicitly accounted for. Notably, this work features numerous instances where predictions have led to the re-measurement and amendment of erroneous experimental pKa values, i.e., theory has corrected experiment. For each of the four ionizable group case studies that are featured (guanidines, sulfonamides, 1,3-diketones and benzoic/naphthoic acids), in addition to the derivation and validation of predictive equations, a rationale is also presented to explain why and how the AIBL-pKa relationship occurs.

Additional content not available electronically

Supporting Information available as .pdf

Keyword(s)

QSPR; drug design; pKa prediction; physical organic chemistry; property prediction; theoretical chemistry

Bibliographic metadata

Institutional metadata

Record metadata