Accurate Lattice Free Energies of Packing Polymorphs from Probabilistic Generative Models

Journal of Chemical Theory and Computation, Год журнала: 2025, Номер unknown

Опубликована: Фев. 21, 2025

Finite-temperature lattice free energy differences between polymorphs of molecular crystals are fundamental to understanding and predicting the relative stability relationships underpinning polymorphism, yet computationally expensive obtain. Here, we implement critically assess machine-learning-enabled targeted calculations derived from flow-based generative models compute difference two ice crystal (Ice XI Ic), modeled with a fully flexible empirical classical force field. We demonstrate that even when remapping an analytical reference distribution, such methods enable cost-effective accurate calculation disconnected metastable ensembles trained on locally ergodic data sampled exclusively interest. Unlike perturbation methods, as Einstein method, approach analyzed in this work requires no additional sampling intermediate perturbed Hamiltonians, offering significant computational savings. To systematically accuracy monitored convergence estimates during training by implementing overfitting-aware weighted averaging strategy. By comparing our results ground-truth computed efficiency different model architectures, employing representations supercell degrees freedom (Cartesian vs quaternion-based). conduct assessment supercells sizes temperatures assessing extrapolating energies thermodynamic limit. While at low small system sizes, perform similar accuracy. note for larger systems high temperatures, choice representation is key obtaining generalizable quality comparable obtained method. believe be stepping stone toward efficient larger, more complex crystals.

Язык: Английский

---

Machine learning in molecular biophysics: Protein allostery, multi-level free energy simulations, and lipid phase transitions

Biophysics Reviews, Год журнала: 2025, Номер 6(1)

Опубликована: Фев. 12, 2025

Machine learning (ML) techniques have been making major impacts on all areas of science and engineering, including biophysics. In this review, we discuss several applications ML to biophysical problems based our recent research. The topics include the use identify hotspot residues in allosteric proteins using deep mutational scanning data analyze how mutations these hotspots perturb co-operativity framework a statistical thermodynamic model, improve accuracy free energy simulations by integrating from different levels potential functions, determine phase transition temperature lipid membranes. Through examples, illustrate unique value extracting patterns or parameters complex sets, as well remaining limitations. By implementing approaches context physically motivated models computational frameworks, are able gain deeper mechanistic understanding better convergence numerical simulations. We conclude briefly discussing introduced can be further expanded tackle more problems.

Язык: Английский

---

Thermodynamic Interpolation: A Generative Approach to Molecular Thermodynamics and Kinetics

Journal of Chemical Theory and Computation, Год журнала: 2025, Номер unknown

Опубликована: Фев. 23, 2025

Using normalizing flows and reweighting, Boltzmann generators enable equilibrium sampling from a distribution, defined by an energy function thermodynamic state. In this work, we introduce interpolation (TI), which allows for generating statistics in temperature-controllable way. We TI flavors that work directly the ambient configurational space, mapping between different states or through latent, normally distributed reference Our ambient-space approach specification of arbitrary target temperatures, ensuring generalizability within temperature range training set demonstrating potential extrapolation beyond it. validate effectiveness on model systems exhibit metastability nontrivial dependencies. Finally, demonstrate how to combine TI-based estimate free differences various perturbation methods provide corresponding approximated kinetic rates, estimated generator extended dynamic mode decomposition (gEDMD).

Язык: Английский

---

Learning mappings between equilibrium states of liquid systems using normalizing flows

The Journal of Chemical Physics, Год журнала: 2025, Номер 162(18)

Опубликована: Май 8, 2025

Generative models and, in particular, normalizing flows are a promising tool statistical mechanics to address the sampling problem condensed-matter systems. In this work, we investigate potential of learn transformation map different liquid systems into each other while allowing at same time obtain an unbiased equilibrium distribution. We apply methodology mapping small system fully repulsive disks modeled via Weeks–Chandler–Andersen Lennard-Jones phase coordinates diagram. improvement relative effective sample size generated distribution up factor six compared direct reweighting. show that can have strong dependency on thermodynamic parameters source and target system.

Язык: Английский

---