The Rosetta all-atom energy function for macromolecular modeling and design

By Rebecca F. Alford, Andrew Leaver-Fay, J. Jeliazkov, Matthew J. O’Meara, Frank DiMaio, Hahnbeom Park, Maxim V. Shapovalov, P. Renfrew, Vikram Khipple Mulligan, K. Kappel and 10 more. 2017

1,486 citations recorded by the source corpus.

Over the past decade, the Rosetta biomolecular modeling suite has informed diverse biological questions and engineering challenges ranging from interpretation of low-resolution structural data to design of nanomaterials, protein therapeutics, and vaccines. Central to Rosetta’s success is the energy function: amodel parameterized from small molecule and X-ray crystal structure data used to approximate the energy associated with each biomolecule conformation. This paper describes the mathematical models and physical concepts that underlie the latest Rosetta energy function, beta_nov15. Applying these concepts,we explain how to use Rosetta energies to identify and analyze the features of biomolecular models.Finally, we discuss the latest advances in the energy function that extend capabilities from soluble proteins to also include membrane proteins, peptides containing non-canonical amino acids, carbohydrates, nucleic acids, and other macromolecules.

Subfields: Materials Chemistry, Molecular Biology, Computational Chemistry & Drug Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Materials Chemistry

In Molecular Biology

In Computational Chemistry & Drug Discovery

Citations in the corpus

Generalized Biomolecular Modeling and Design with RoseTTAFold All-Atom (2023, positive): “We found that predictions for protein – small molecule complexes in our recent PDB set with lower computedbindingenergies(Rosetta D G )( 32 , 33 ) were more accurate (Fig.”
Protein Structure and Sequence Generation with Equivariant Denoising Diffusion Probabilistic Models (2022, background): “However, for the task of structure generation, stochastic search algorithms based on handcrafted energy functions and heuristic sampling approaches are still in wide use [25, 34, 2, 26].”
Unified rational protein engineering with sequence-based deep representation learning (2019, methodology): “…compared to typical protein engineering tools, we obtained the published Rosetta total energy estimates with the ‘beta_nov15’ version of the energy function 67,68 and buried nonpolar surface area of the designed structure for the stability prediction for de novo designed mini proteins dataset 5 .”
Learning inverse folding from millions of predicted structures (2022, methodology): “The most well-established approaches to this problem use an energy function which directly models the physical basis of a protein’s folded state (Alford et al., 2017).”
Protein sequence design with a learned potential (2022, methodology): “…commonly involve specifying an energy function and sampling sequence space to ﬁ nd a minimum-energy con ﬁ guration 13 – 15 , and enormous effort has gone into the development of carefully modeled and parameterized energy functions to guide design, which continue to be iteratively re ﬁ ned 16,17 .”
Protein sequence design by conformational landscape optimization (2021, methodology): “Backbones were constructed using blueprints (29) with distance and angle constraints to guide the formation of β-sheet backbone hydrogen bonds (30), followed by sequence design with Rosetta [FastDesign (28, 31) in conjunction with LayerDesign (21) and using the Rosetta all-atom energy function (32) ”
Low-N protein engineering with data-efficient deep learning (2020, background): “Additionally, they often require expert judgment to learn from data, which may include modifying energy functions for biophysical design 61 , and iteratively designing and testing structure-guided mutation combinations 19,62–65 .”
Energy-based models for atomic-resolution protein conformations (2020, methodology): “Rosetta, for example, uses a combination of physically motivated terms and knowledge-based potentials (Alford et al., 2017) to model proteins and other macromolecules.”
Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants (2021, methodology): “N eff score is defined as the number of clusters after the raw MSA inputs were clustered at the 62% sequence identity using CD-HIT 66 with the word length of 4, as used previously.”
Pairing a high-resolution statistical potential with a nucleobase-centric sampling algorithm for improving RNA model refinement (2021, methodology): “It employed a Rosetta all-atom force ﬁ eld that mixes physical and knowledge-based scores with RNA-speci ﬁ c terms for Watson – Crick (WC) base pairing, base stacking, and torsional potentials 25 .”
Flexible backbone assembly and refinement of symmetrical homomeric complexes (2018, methodology): “The Rosetta all-atom score function (Alford et al., 2017; Park et al., 2016) scores models close to the native conformation more favorably than non-native models (Figs.”
Accurate prediction of nucleic acid and protein-nucleic acid complexes using RoseTTAFoldNA (2022, methodology): “…and 1556 protein-nucleic acid complex clusters compared to 26128 all protein clusters) , we also incorporated physical information in the form of Lennard-Jones and hydrogen-bonding energies [12] as input features to the final refinement layers, and as part of the loss function during fine-tuning.”