Foundation Model Extends MACE With Electrostatics
On Feb. 23, 2026, Batatia et al. present an electrostatic foundation model that extends the MACE architecture with polarizable iterations and global charge equilibration. Trained on OMol25's 100 million hybrid DFT calculations, it achieves chemical accuracy across benchmarks, including sub-kcal/mol X23-DMC crystal energies and fourfold better protein–ligand interaction predictions. The model handles variable charge and spin states and scales from small molecules to complexes.
Key Points
- 1Introduces electrostatic foundation model extending MACE with polarizable iterations and global charge equilibration
- 2Improves noncovalent and supramolecular predictions, sub-kcal/mol crystal energies and fourfold protein-ligand gains
- 3Enables efficient, accurate modeling of variable charge/spin systems for drug discovery and materials
Scoring Rationale
High novelty and broad applicability across chemistry and drug discovery, limited by preprint status and need for independent validation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
