Deep Learning in Protein Structural Modeling and Design

Patterns, Journal Year: 2020, Volume and Issue: 1(9), P. 100142 - 100142

Published: Nov. 12, 2020

Deep learning is catalyzing a scientific revolution fueled by big data, accessible toolkits, and powerful computational resources, impacting many fields, including protein structural modeling. Protein modeling, such as predicting structure from amino acid sequence evolutionary information, designing proteins toward desirable functionality, or properties behavior of protein, critical to understand engineer biological systems at the molecular level. In this review, we summarize recent advances in applying deep techniques tackle problems modeling design. We dissect emerging approaches using for discuss challenges that must be addressed. argue central importance structure, following "sequence → function" paradigm. This review directed help both biologists gain familiarity with methods applied computer scientists perspective on biologically meaningful may benefit techniques.

Language: Английский

---

Development and Benchmarking of Open Force Field 2.0.0: The Sage Small Molecule Force Field

Journal of Chemical Theory and Computation, Journal Year: 2023, Volume and Issue: 19(11), P. 3251 - 3275

Published: May 11, 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 work, 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 demonstrates estimating physical properties, including comparison experimental from various thermodynamic databases properties ΔHmix, ρ(x), ΔGsolv, ΔGtrans. Additionally, benchmarked binding free energies (ΔGbind), where yields results statistically similar All made publicly available along complete details reproduce training at https://github.com/openforcefield/openff-sage.

Language: Английский

---

Open-Source Machine Learning in Computational Chemistry

Journal of Chemical Information and Modeling, Journal Year: 2023, Volume and Issue: 63(15), P. 4505 - 4532

Published: July 19, 2023

The field of computational chemistry has seen a significant increase in the integration machine learning concepts and algorithms. In this Perspective, we surveyed 179 open-source software projects, with corresponding peer-reviewed papers published within last 5 years, to better understand topics being investigated by approaches. For each project, provide short description, link code, accompanying license type, whether training data resulting models are made publicly available. Based on those deposited GitHub repositories, most popular employed Python libraries identified. We hope that survey will serve as resource learn about or specific architectures thereof identifying accessible codes topic basis. To end, also include for generating fundamental learning. our observations considering three pillars collaborative work, open data, source (code), models, some suggestions community.

Language: Английский

---

EspalomaCharge: Machine Learning-Enabled Ultrafast Partial Charge Assignment

The Journal of Physical Chemistry A, Journal Year: 2024, Volume and Issue: 128(20), P. 4160 - 4167

Published: May 8, 2024

Atomic partial charges are crucial parameters in molecular dynamics simulation, dictating the electrostatic contributions to intermolecular energies and thereby potential energy landscape. Traditionally, assignment of has relied on surrogates ab initio semiempirical quantum chemical methods such as AM1-BCC is expensive for large systems or numbers molecules. We propose a hybrid physical/graph neural network-based approximation widely popular charge model that orders magnitude faster while maintaining accuracy comparable differences implementations. Our approach couples graph network streamlined equilibration order predict molecule-specific atomic electronegativity hardness parameters, followed by analytical determination optimal charge-equilibrated preserve total charge. This scales linearly with number atoms, enabling first time use fully consistent models small molecules biopolymers construction next-generation self-consistent biomolecular force fields. Implemented free open source package EspalomaCharge, this provides drop-in replacements both AmberTools antechamber Open Force Field Toolkit charging workflows, addition stand-alone generation interfaces. Source code available at https://github.com/choderalab/espaloma-charge.

Language: Английский

---

Differentiable simulation to develop molecular dynamics force fields for disordered proteins

Chemical Science, Journal Year: 2024, Volume and Issue: 15(13), P. 4897 - 4909

Published: Jan. 1, 2024

The a99SB- disp force field and GBNeck2 implicit solvent model are improved to better describe disordered proteins. 5 ns differentiable molecular simulations used jointly optimise 108 parameters match explicit trajectories.

Language: Английский

---

Machine-learned molecular mechanics force fields from large-scale quantum chemical data

Chemical Science, Journal Year: 2024, Volume and Issue: 15(32), P. 12861 - 12878

Published: Jan. 1, 2024

A generalized and extensible machine-learned molecular mechanics force field trained on over 1.1 million QC data applicable for drug discovery applications. Figure reproduced from the arXiv:201001196 preprint under arXiv non-exclusive license.

Language: Английский

---

The emergence of machine learning force fields in drug design

Medicinal Research Reviews, Journal Year: 2024, Volume and Issue: 44(3), P. 1147 - 1182

Published: Jan. 3, 2024

In the field of molecular simulation for drug design, traditional mechanic force fields and quantum chemical theories have been instrumental but limited in terms scalability computational efficiency. To overcome these limitations, machine learning (MLFFs) emerged as a powerful tool capable balancing accuracy with MLFFs rely on relationship between structures potential energy, bypassing need preconceived notion interaction representations. Their depends models used, quality volume training data sets. With recent advances equivariant neural networks high-quality datasets, significantly improved their performance. This review explores MLFFs, emphasizing design. It elucidates MLFF principles, provides development validation guidelines, highlights successful implementations. also addresses challenges developing applying MLFFs. The concludes by illuminating path ahead outlining to be opportunities harnessed. inspires researchers embrace investigations new perform simulations

Language: Английский

---

Application of Modern Artificial Intelligence Techniques in the Development of Organic Molecular Force Fields

Physical Chemistry Chemical Physics, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

The molecular force field (FF) determines the accuracy of dynamics (MD) and is one major bottlenecks that limits application MD in design. Recently, artificial intelligence (AI) techniques, such as machine-learning potentials (MLPs), have been rapidly reshaping landscape MD. Meanwhile, organic systems feature unique characteristics, require more careful treatment both model construction, optimization, validation. While an accurate generic still missing, significant progress has made with facilitation AI, warranting a promising future. In this review, we provide overview various types AI techniques used FF development discuss advantages weaknesses these methodologies. We show how methods unprecedented capabilities many tasks potential fitting, atom typification, automatic optimization. it also worth noting efforts are needed to improve transferability model, develop comprehensive database, establish standardized validation procedures. With discussions, hope inspire solve existing problems, eventually leading birth next-generation FFs.

Language: Английский

---

Grappa – a machine learned molecular mechanics force field

Chemical Science, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

We propose Grappa, a machine learned molecular mechanics force field for proteins. operating on the graph, accurately predicts energies and forces agrees with experimental data such as J -couplings folding free energies.

Language: Английский

---

Scaling Graph Neural Networks to Large Proteins

Journal of Chemical Theory and Computation, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 6, 2025

Graph neural network (GNN) architectures have emerged as promising force field models, exhibiting high accuracy in predicting complex energies and forces based on atomic identities Cartesian coordinates. To expand the applicability of GNNs, machine learning fields more broadly, optimizing their computational efficiency is critical, especially for large biomolecular systems classical molecular dynamics simulations. In this study, we address key challenges existing GNN benchmarks by introducing a dataset, DISPEF, which comprises large, biologically relevant proteins. DISPEF includes 207,454 proteins with sizes up to 12,499 atoms features diverse chemical environments, spanning folded disordered regions. The implicit solvation free energies, used training targets, represent particularly challenging case due many-body nature, providing stringent test evaluating expressiveness models. We benchmark performance seven GNNs emphasizing importance directly accounting long-range interactions enhance model transferability. Additionally, present novel multiscale architecture, termed Schake, delivers transferable computationally efficient energy predictions Our findings offer valuable insights tools advancing protein modeling applications.

Language: Английский

---