Abstract

The acid dissociation constant (pKa) is the key parameter to define the extent of ionization of a drug molecule and is used for ADME properties evaluation via the pH-dependent distribution coefficient, logD. We present a method for pKa prediction using a predefined reference database and structural fingerprints based on a multilevel neighborhoods description of the ionizable atom(s). This database mining approach is suitable for screening large compound collections for HTS compound prioritization and external compound acquisition. In addition to pKa prediction it provides medicinal chemists rapid access to already available pKa measurements and hints for manipulating the chemical structure to increase or decrease pKa.

Keywords: pKa prediction, multilevel neighborhoods of atoms, database mining, cheminformatics