“…Several pK a prediction programs exist (Liao and Nicklaus, 2009). Commercial predictors span a range of mechanisms to predict the protonation state of particular atoms, including linear free energy relationships (LFER) that use a dictionary of chemical substructures (Lee et al, 2007), quantitative structure-property relationships (QSPR) (Jover et al, 2008;Palaz et al, 2012), and quantum chemical and ab initio methods (Bochevarov et al, 2013;Eckert and Klamt, 2006;Eckert et al, 2009;Klamt et al, 2010;Klamt et al, 2003;Vareková et al, 2011). Semi-empirical models calculate descriptors for each ionizable chemical functional group, after which pK a values are predicted using machine learning or tree-based models (Jelfs et al, 2007;Xing et al, 2003).…”