MF-SuP-pK(a): Multi-fidelity modeling with subgraph pooling mechanism for pK(a) prediction

Acid-base dissociation constant (pK(a)) is a key physicochemical parameter in chemical science, especially in organic synthesis and drug discovery. Current methodologies for pK(a) prediction still suffer from limited applicability domain and lack of chemical insight. Here we present MF-SuP-pK(a) (multi-fidelity modeling with subgraph pooling for pK(a) prediction), a novel pK(a) prediction model that utilizes subgraph pooling, multi-fidelity learning and data augmentation. In our model, a knowledge-aware subgraph pooling strategy was designed to capture the local and global environments around the ionization sites for micro-pK(a) prediction. To overcome the scarcity of accurate pK(a) data, low-fidelity data (computational pK(a)) was used to fit the high-fidelity data (experimental pK(a)) through transfer learning. The final MF-SuP-pK(a) model was constructed by pre-training on the augmented ChEMBL data set and fine-tuning on the DataWarrior data set. Extensive evaluation on the DataWarrior data set and three benchmark data sets shows that MF-SuP-pK(a) achieves superior performances to the state-of-the-art pK(a) prediction models while requires much less high-fidelity training data. Compared with Attentive FP, MF-SuP-pK(a) achieves 23.83% and 20.12% improvement in terms of mean absolute error (MAE) on the acidic and basic sets, respectively.