Overview of the SAMPL6 pK_aChallenge: Evaluating small molecule microscopic and macroscopic pK_apredictions

Abstract

The prediction of acid dissociation constants (pKa) is a prerequisite for predicting many other properties of a small molecule, such as its protein-ligand binding affinity, distribution coefficient (logD), membrane permeability, and solubility. The prediction of each of these properties requires knowledge of the relevant protonation states and solution free energy penalties of each state. The SAMPL6 pKa Challenge was the first time that a separate challenge was conducted for evaluating pKa predictions as part of the Statistical Assessment of Modeling of Proteins and Ligands (SAMPL) exercises. This challenge was motivated by significant inaccuracies observed in prior physical property prediction challenges, such as the SAMPL5 logD Challenge, caused by protonation state and pKa prediction issues. The goal of the pKa challenge was to assess the performance of contemporary pKa prediction methods for drug-like molecules. The challenge set was composed of 24 small molecules that resembled fragments of kinase inhibitors, a number of which were multiprotic. Eleven research groups contributed blind predictions for a total of 37 pKa distinct prediction methods. In addition to blinded submissions, four widely used pKa prediction methods were included in the analysis as reference methods. Collecting both microscopic and macroscopic pKa predictions allowed in-depth evaluation of pKa prediction performance. This article highlights deficiencies of typical pKa prediction evaluation approaches when the distinction between microscopic and macroscopic pKas is ignored; in particular, we suggest more stringent evaluation criteria for microscopic and macroscopic pKa predictions guided by the available experimental data. Top-performing submissions for macroscopic pKa predictions achieved RMSE of 0.7-1.0 pKa units and included both quantum chemical and empirical approaches, where the total number of extra or missing macroscopic pKas predicted by these submissions were fewer than 8 for 24 molecules. A large number of submissions had RMSE spanning 1-3 pKa units. Molecules with sulfur-containing heterocycles or iodo and bromo groups were less accurately predicted on average considering all methods evaluated. For a subset of molecules, we utilized experimentally-determined microstates based on NMR to evaluate the dominant tautomer predictions for each macroscopic state. Prediction of dominant tautomers was a major source of error for microscopic pKa predictions, especially errors in charged tautomers. The degree of inaccuracy in pKa predictions observed in this challenge is detrimental to the protein-ligand binding affinity predictions due to errors in dominant protonation state predictions and the calculation of free energy corrections for multiple protonation states. Underestimation of ligand pKa by 1 unit can lead to errors in binding free energy errors up to 1.2 kcal/mol. The SAMPL6 pKa Challenge demonstrated the need for improving pKa prediction methods for drug-like molecules, especially for challenging moieties and multiprotic molecules.

SAMPL6 Challenge Results from pKa Predictions Based on a General Gaussian Process Model

“…Macroscopic pK a s are then analytically calculated from a combination of microscopic transitions (Section 2.4). We used our model to predict microscopic (type I) and macroscopic (type III) pK a s for 24 drug like molecules in the SAMPL6 blind challenge [13].…”

“…For each molecule, the experimental values were matched to microscopic predictions that resulted in the lowest error, using the same closest and Hungarian pairing algorithms described in Section 3.0.1. These results are available for all submissions on the SAMPL6 GitHub repository [13] However, here, we have chosen to focus on analysis of directly measured microscopic predictions rather than microscopic predictions with presumed correspondence to macroscopic measurements. While many macroscopic transitions are likely dominated by a single microscopic transition, including those in section 3.1.1, we cannot know for sure that this is true for every molecule in this set.…”

“…Commercial models provide a reference for more of our microscopic pK a predictions-In an attempt to evaluate a wider range of microscopic pK a s, we compared our predictions with some of the top results from the macroscopic analysis. SimulationsPlus' pK a Predictor [20] (submission ID hdiyq) and ACD Lab's pK a GALAS [49] (submission ID 8qph) both performed better than our approach in the macroscopic pK a challenge compared to experiment and provided type I predictions [13]. We will refer to these two submissions as S+ and ACD respectively.…”

“…For each type of submission participants were encouraged, but not required, to submit all predictions their model generated for every molecule. SAMPL6 organizers then evaluated predictions based on experimental results for all macroscopic and a select set of microscopic pK a s [13]. Details for the challenge including experimental results, all submitted predictions, and an overview analysis are available online (github.com/MobleyLab/SAMPL6).…”