MolGpka: A Web Server for Small Molecule pKa Prediction Using a Graph-Convolutional Neural Network

pK(a) is an important property in the lead optimization process since the charge state of a molecule in physiologic pH plays a critical role in its biological activity, solubility, membrane permeability, metabolism, and toxicity. Accurate and fast estimation of small molecule pK(a) is vital during the drug discovery process. We present MolGpKa, a web server for pK(a) prediction using a graph-convolutional neural network model. The model works by learning pK(a) related chemical patterns automatically and building reliable predictors with learned features. ACD/pK(a) data for 1.6 million compounds from the ChEMBL database was used for model training. We found that the performance of the model is better than machine learning models built with human-engineered fingerprints. Detailed analysis shows that the substitution effect on pK(a) is well learned by the model. MolGpKa is a handy tool for the rapid estimation of pK(a) during the ligand design process. The MolGpKa server is freely available to researchers and can be accessed at https://xundrug.cn/molgpka.

Automated summary

pK(a) is a crucial property in lead optimization and drug discovery processes, and MolGpKa is a web server that utilizes a graph-convolutional neural network model for pK(a) prediction. The model shows better performance than machine learning models with human-engineered fingerprints, and it effectively learns the substitution effect on pK(a). MolGpKa is a convenient tool for rapid estimation of pK(a) during ligand design.