KaMLs for Predicting Protein pKa Values and Ionization States: Are Trees All You Need?

Abstract

Despite its importance in understanding biology and computer-aided drug discovery, the accurate prediction of protein ionization states remains a formidable challenge. Physics-based approaches struggle to capture the small, competing contributions in the complex protein environment, while machine learning (ML) is hampered by scarcity of experimental data. Here we report the development of pKa ML (KaML) models based on decision trees and graph attention networks (GAT), exploiting physicochemical understanding and a new experiment pKa database (PKAD-3) enriched with highly shifted pKas. KaML-CBtree significantly outperforms the current state of the art in predicting pKa values and ionization states across all six titratable amino acids, notably achieving accurate predictions for deprotonated cysteines and lysines - a blind spot in previous models. The superior performance of KaMLs is achieved in part through several innovations, including separate treatment of acid and base, data augmentation using AlphaFold structures, and model pretraining on a theoretical pKa database. We also introduce the classification of protonation states as a metric for evaluating pKa prediction models. A meta-feature analysis suggests a possible reason for the lightweight tree model to outperform the more complex deep learning GAT. We release an end-to-end pKa predictor based on KaML-CBtree and the new PKAD-3 database, which facilitates a variety of applications and provides the foundation for further advances in protein electrostatics research. TOC Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=111 SRC="FIGDIR/small/622800v3_ufig1.gif" ALT="Figure 1"> View larger version (16K): org.highwire.dtl.DTLVardef@15d2256org.highwire.dtl.DTLVardef@1795151org.highwire.dtl.DTLVardef@1c9ae02org.highwire.dtl.DTLVardef@1bf17a9_HPS_FORMAT_FIGEXP M_FIG C_FIG