Regularized by Physics: Graph Neural Network Parametrized Potentials for the Description of Intermolecular Interactions

Journal of Chemical Theory and Computation, Journal Year: 2023, Volume and Issue: 19(2), P. 562 - 579

Published: Jan. 12, 2023

Simulations of molecular systems using electronic structure methods are still not feasible for many biological importance. As a result, empirical such as force fields (FF) have become an established tool the simulation large and complex systems. The parametrization FF is, however, time-consuming has traditionally been based on experimental data. Recent years therefore seen increasing efforts to automatize or replace with machine-learning (ML) potentials. Here, we propose alternative strategy parametrize FF, which makes use ML gradient-descent optimization while retaining functional form founded in physics. Using predefined is shown enable interpretability, robustness, efficient simulations over long time scales. To demonstrate strength proposed method, fixed-charge polarizable model trained ab initio potential-energy surfaces. Given only information about constituting elements, topology, reference potential energies, models successfully learn assign atom types corresponding parameters from scratch. resulting validated wide range experimentally computationally derived properties including dimers, pure liquids, crystals.

Language: Английский

---

Advancing Force Fields Parameterization: A Directed Graph Attention Networks Approach

Journal of Chemical Theory and Computation, Journal Year: 2024, Volume and Issue: 20(13), P. 5558 - 5569

Published: June 14, 2024

Force fields (FFs) are an established tool for simulating large and complex molecular systems. However, parametrizing FFs is a challenging time-consuming task that relies on empirical heuristics, experimental data, computational data. Recent efforts aim to automate the assignment of FF parameters using pre-existing databases on-the-fly

Language: Английский

---

The Open Force Field Initiative: Open Software and Open Science for Molecular Modeling

The Journal of Physical Chemistry B, Journal Year: 2024, Volume and Issue: unknown

Published: July 11, 2024

Force fields are a key component of physics-based molecular modeling, describing the energies and forces in system as function positions atoms molecules involved. Here, we provide review scientific status report on work Open Field (OpenFF) Initiative, which focuses science, infrastructure data required to build next generation biomolecular force fields. We introduce OpenFF Initiative related Consortium, describe its approach field development software, discuss accomplishments date well future plans. releases both software under open permissive licensing agreements enable rapid application, validation, extension, modification tools. lessons learned this new development. also highlight ways that other researchers can get involved, some recent successes outside taking advantage tools data.

Language: Английский

---

Substituting density functional theory in reaction barrier calculations for hydrogen atom transfer in proteins

Chemical Science, Journal Year: 2024, Volume and Issue: 15(7), P. 2518 - 2527

Published: Jan. 1, 2024

Hydrogen atom transfer (HAT) reactions are important in many biological systems. As these hard to observe experimentally, it is of high interest shed light on them using simulations. Here, we present a machine learning model based graph neural networks for the prediction energy barriers HAT proteins. input, uses exclusively non-optimized structures as obtained from classical It was trained more than 17 000 calculated hybrid density functional theory. We built and evaluated context collagen, but show that same workflow can easily be applied other or synthetic polymers. obtain relevant (small reaction distances) with good predictive power (R2 ∼ 0.9 mean absolute error <3 kcal mol-1). inference speed high, this enables evaluations dozens chemical situations within seconds. When combined molecular dynamics kinetic Monte-Carlo scheme, paves way toward reactive

Language: Английский

---

Computational Chemistry as Applied in Environmental Research: Opportunities and Challenges

ACS ES&T Engineering, Journal Year: 2023, Volume and Issue: 4(1), P. 66 - 95

Published: Oct. 12, 2023

The constant development of computer systems and infrastructure has allowed computational chemistry to become an important component environmental research. In the past decade, application quantum classical mechanical calculations model understand increased exponentially. this review, we highlight various applications techniques in areas research (e.g., wastewater/air treatment, sensing, biodegradation). We briefly describe each approach, starting with principle methods followed by molecular mechanics (MM), dynamics (MD), hybrid QM/MM methods. recent introduction artificial intelligence machine learning their potential disrupt field are also discussed. Challenges current future directions address them presented.

Language: Английский

---

Development and Benchmarking of Open Force Field 2.0.0 — the Sage Small Molecule Force Field

Published: Jan. 10, 2023

We introduce the Open Force Field (OpenFF) 2.0.0 small molecule force field for drug-like molecules, code-named Sage, which builds upon our previous iteration, Parsley. OpenFF fields are based on direct chemical perception, generalizes easily to highly diverse sets of chemistries substructure queries. Like iterations, Sage generation was validated in protein-ligand simulations be compatible with AMBER biopolymer fields. In this paper we detail methodology used develop field, as well innovations and improvements introduced since release Parsley 1.0.0. One particularly significant feature is a set improved Lennard-Jones (LJ) parameters retrained against condensed phase mixture data, first refit LJ line. also includes valence larger database quantum calculations than versions, how fitting performed. benchmarks show general metrics performance chemistry reference data such root mean square deviations (RMSD) optimized conformer geometries, torsion fingerprint (TFD), relative energetics (ΔΔ𝐸). present variety these some cases other biomolecular demonstrates estimating physical properties, including comparison experimental from various thermodynamic databases properties Δ𝐻_𝑚𝑖𝑥, ρ(𝑥), Δ𝐺_𝑠𝑜𝑙𝑣 Δ𝐺_𝑡𝑟𝑎𝑛𝑠. Additionally, benchmarked binding free energies (Δ𝐺_𝑏𝑖𝑛𝑑), where yields results statistically similar All made publicly available along complete details reproduce training at https://github.com/openforcefield/openff-sage.

Language: Английский

---

Evolutionary-docking targeting bacterial FtsZ

Published: March 7, 2023

To accurately predict binding of inhibitors to the FtsZ cell division protein antibiotic-resistance Staphilococcus aureus pathogen, evolutionary library docking, ligand-efficiency predictions and rank consensus docking strategies have been sequentially applied. Starting from crystallographic bound model PC190723 reference ligand, fragments were derived generate children molecules fitting low docking-scores with molecular sizes hydrophobicities using DataWarrior Build Evolutionary Library. fragment combined toxicity filters, ranks ligand efficiencies AutoDockVina identified new benzamide non-benzamide chemotypes nanomolar improved specificities continue anti-FtsZ investigations

Language: Английский

---

Machine-learned molecular mechanics force field for the simulation of protein-ligand systems and beyond

arXiv (Cornell University), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 1, 2023

The development of reliable and extensible molecular mechanics (MM) force fields -- fast, empirical models characterizing the potential energy surface systems is indispensable for biomolecular simulation computer-aided drug design. Here, we introduce a generalized machine-learned MM field, \texttt{espaloma-0.3}, an end-to-end differentiable framework using graph neural networks to overcome limitations traditional rule-based methods. Trained in single GPU-day fit large diverse quantum chemical dataset over 1.1M calculations, \texttt{espaloma-0.3} reproduces energetic properties domains highly relevant discovery, including small molecules, peptides, nucleic acids. Moreover, this field maintains energy-minimized geometries molecules preserves condensed phase self-consistently parametrizing proteins ligands produce stable simulations leading accurate predictions binding free energies. This methodology demonstrates significant promise as path forward systematically building more that are easily new interest.

Language: Английский

---

Incorporating Neural Networks into the AMOEBA Polarizable Force Field

The Journal of Physical Chemistry B, Journal Year: 2024, Volume and Issue: 128(10), P. 2381 - 2388

Published: March 6, 2024

Neural network potentials (NNPs) offer significant promise to bridge the gap between accuracy of quantum mechanics and efficiency molecular in simulation. Most NNPs rely on locality assumption that ensures model's transferability scalability thus lack treatment long-range interactions, which are essential for systems condensed phase. Here we present an integrated hybrid model, AMOEBA+NN, combines AMOEBA potential short- noncovalent atomic interactions NNP capture remaining local covalent contributions. The AMOEBA+NN model was trained conformational energy ANI-1x data set tested several external sets ranging from small molecules tetrapeptides. demonstrated substantial improvements over baseline models term as molecule size increased, suggesting its a next-generation approach chemically accurate simulations.

Language: Английский

---

Fluorinated Protein–Ligand Complexes: A Computational Perspective

The Journal of Physical Chemistry B, Journal Year: 2024, Volume and Issue: 128(25), P. 5925 - 5934

Published: June 17, 2024

Fluorine is an element renowned for its unique properties. Its powerful capability to modulate molecular properties makes it attractive substituent protein binding ligands; however, the rational design of fluorination can be challenging with effects on interactions and energies being difficult predict. In this Perspective, we highlight how computational methods help us understand role fluorine in protein–ligand a focus simulation. We underline importance accurate force field, present fluoride channels as showcase biomolecular fluorine, discuss specific like ability form hydrogen bonds aryl groups. put special emphasis disruption water networks entropic effects.

Language: Английский

---