article Machine Learning Potentials Experiment Automation Drug Discovery Materials Informatics

Discovery of Novel Celecoxib Polymorphs Using AIMNet2 Machine Learning Interatomic Potential

Peikun Zheng, Yuriy Abramov, Changquan Calvin Sun, Olexandr Isayev

2025

Highlight

Polymorphism plays a pivotal role in defining the solid-state properties of pharmaceutical compounds, yet the discovery and accurate energy ranking of polymorphs remain a challenge.

Abstract

Polymorphism plays a pivotal role in defining the solid-state properties of pharmaceutical compounds, yet the discovery and accurate energy ranking of polymorphs remain a challenge. Here, we leverage a fine-tuned machine-learned interatomic potential AIMNet2 to explore the polymorphic landscape of celecoxib, a clinically important COX-2 inhibitor. Our approach combines GPU-accelerated crystal structure generation, active learning-guided model refinement, and quasi-harmonic free-energy corrections. The workflow successfully reproduces the experimental energy hierarchy of known polymorphs and identifies several novel low-energy structures with distinct packing motifs. In addition, we evaluate the elastic properties and thermal expansion effects across polymorphs, revealing structural features that underpin mechanical flexibility and thermodynamic preferences. This study demonstrates the power of AIMNet2-based crystal structure prediction for resolving complex pharmaceutical polymorphism and offers a powerful tool for future polymorph discovery and solid-state optimization.

Keywords

Cite This Paper

@article{Zheng2025,
  author = {Zheng, Peikun and Abramov, Yuriy and Sun, Changquan Calvin and Isayev, Olexandr},
  title = {Discovery of Novel Celecoxib Polymorphs Using AIMNet2 Machine Learning Interatomic Potential},
  year = {2025},
  doi = {10.26434/chemrxiv-2025-nhmr1},
  url = {http://dx.doi.org/10.26434/chemrxiv-2025-nhmr1},
  publisher = {American Chemical Society (ACS)},
  keywords = {machine learning, active learning},
  researchAreas = {ml-potentials, experiment-automation, drug-discovery, materials-informatics},
  highlight = {Polymorphism plays a pivotal role in defining the solid-state properties of pharmaceutical compounds, yet the discovery and accurate energy ranking of polymorphs remain a challenge.}
}

Related Research Areas

Related Publications

2025
cited 3

Transferable Machine Learning Interatomic Potential for Pd-Catalyzed Cross-Coupling Reactions

Anstine D., Zubatyuk R., Gallegos L., Paton R., Wiest O., Nebgen B., Jones T., Gomes G., Tretiak S., Isayev O.

Machine Learning Potentials Reactions & Reactivity Experiment Automation Materials Informatics

Finding efficient substrate-catalyst combinations for palladium-catalyzed cross-coupling reactions remains a critical challenge in synthetic chemistry, with broad implications for pharmaceutical and materials manufacturing.

DOI
2020
cited 212

The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules

Smith J. S., Zubatyuk R., Nebgen B., Lubbers N., Barros K., Roitberg A. E., Isayev O., Tretiak S.

Scientific Data 7

Machine Learning Potentials Quantum Chemistry Experiment Automation

Abstract Maximum diversification of data is a central theme in building generalized and accurate machine learning (ML) models.

DOI
2024
cited 85

Exploring the frontiers of condensed-phase chemistry with a general reactive machine learning potential

Zhang S., Makoś M. Z., Jadrich R. B., Kraka E., Barros K., Nebgen B. T., Tretiak S., Isayev O., Lubbers N., Messerly R. A., et al.

Nature Chemistry 16 , 727–734

Machine Learning Potentials Experiment Automation

Abstract Atomistic simulation has a broad range of applications from drug design to materials discovery.

DOI
2021
cited 94

Teaching a neural network to attach and detach electrons from molecules

Zubatyuk R., Smith J. S., Nebgen B. T., Tretiak S., Isayev O.

Nature Communications 12

Machine Learning Potentials Reactions & Reactivity Drug Discovery Quantum Chemistry

Abstract Interatomic potentials derived with Machine Learning algorithms such as Deep-Neural Networks (DNNs), achieve the accuracy of high-fidelity quantum mechanical (QM) methods in areas traditionally dominated by empirical force fields and allow performing massive simulations.

DOI
2020
cited 63

Towards chemical accuracy for alchemical free energy calculations with hybrid physics-based machine learning / molecular mechanics potentials

Rufa D. A., Bruce Macdonald H. E., Fass J., Wieder M., Grinaway P. B., Roitberg A. E., Isayev O., Chodera J. D.

Drug Discovery Machine Learning Potentials

Towards chemical accuracy for alchemical free energy calculations with hybrid physics-based machine learning / molecular mechanics potentials.

DOI