Scaffolding protein functional sites using deep learning
Open accessPublished version of a preprint
The DOI matches its Crossref record; 104 of the 106 references it cites were found in a registry or an index, and 2 are not checked yet; 21 of the 39 sources its claims cite were checked for support.
Abstract
The binding and catalytic functions of proteins are generally mediated by a small number of functional residues held in place by the overall protein structure. Here, we describe deep learning approaches for scaffolding such functional sites without needing to prespecify the fold or secondary structure of the scaffold. The first approach, “constrained hallucination,” optimizes sequences such that their predicted structures contain the desired functional site. The second approach, “inpainting,” starts from the functional site and fills in additional sequence and structure to create a viable protein scaffold in a single forward pass through a specifically trained RoseTTAFold network. We use these two methods to design candidate immunogens, receptor traps, metalloproteins, enzymes, and protein-binding proteins and validate the designs using a combination of in silico and experimental tests.
Verified 2026-10-05 · 2 references not checked yet
Overview
CoChat Summary
TL;DR
The paper introduces two deep learning methods, constrained hallucination and inpainting, for scaffolding functional sites into de novo protein structures without pre-specifying scaffold fold or secondary structure. It validates designs in silico and experimentally across immunogens, receptor traps, metalloproteins, enzymes, and protein binders.
Key findings
- AlphaFold recapitulated designed structures with overall and motif RMSD typically <2 Å and <1 Å, respectively, and pLDDT >80; inpainting was much faster (1–10 s/design on an NVIDIA RTX2080 GPU) than hallucination (5–20 min/design).
- For RSV-F site V, 37 hallucinated scaffolds were expressed in E. coli and three bound hRSV90 with Kd’s of 0.9–1.3 μM; four inpainted designs bound by yeast display but were poorly expressed in E. coli.
- For the E. coli bacterioferritin di-iron site, 96 inpainted designs were tested: 76 had soluble expression, at least 8 showed a Co2+-binding spectroscopic shift, and 3 had CD spectra consistent with the designed fold and were stabilized by metal binding; mutation of metal-binding residues abolished binding.
- For PD-L1, of 31 designs tested, pdl1_inp_1 bound PD-L1 with KD = 326 nM, compared with HAC PD-1 KD = 110 pM and WT PD-1 KD = 3.9 μM; carbonic anhydrase II active-site designs were <100 residues versus the 261-residue native protein.
Methods
- Constrained hallucination optimized sequences with a composite loss combining a hallucination loss and a motif reconstruction loss using trRosetta and later RosettaFold; inpainting used RFjoint, trained on masked/inpainting sequence and structure recovery, to complete missing sequence and structure in a single forward pass.
- In silico validation used AlphaFold; experimental validation for selected designs included E. coli expression, yeast display, binding assays (Kd, BLI), circular dichroism, metal/terbium fluorescence, and mutagenesis.
- Functional motifs were derived from native proteins including RSV-F sites II and V, hACE2, bacterioferritin di-iron, EF-hand, carbonic anhydrase II, Δ5-3-ketosteroid isomerase, PD-1/PD-L1, p53/Mdm2, and TrkA-NGF.
Limitations
- Stated in the paper: AlphaFold has limited ability to predict protein stability or mutational effects; hallucination is compute-intensive; inpainting usually requires larger input motifs and underperforms hallucination when the missing region is large; some inpainted RSV-F site V designs were poorly expressed in E. coli; di-iron hallucination designs were not tested due to buried polar residues; and side-chain placement for ketosteroid isomerase was difficult, requiring a two-stage approach.
- Evident from the text: many designs were validated only in silico (e.g., receptor traps, enzyme active sites, some Mdm2 binders), and experimental hit rates were often low (e.g., 3/37 RSV-F site V binders, 3/96 di-iron designs with CD/stabilization, 1/31 PD-L1 binders); inpainting also generated less structural diversity than hallucination.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures


![Figure 3. Design of metal binding (A) Di-iron binding site from E. coli cytochrome b1 (1BCF chain A residues 18–25, 27–54, 94–97, 123–130). Colors: native protein scaffold (light yellow); native functional motif (orange); hallucinated scaffold (gray); hallucinated motif (purple); bound metal (blue). Active site residues shown in boxes for di-iron and EF-hand respectively. (B) Absorbance spectra showing of dife_inp_1 (or mutant) in the presence (or not) of an 8-fold molar excess of Co2+. Note the peaks at 520 nm, 555 nm and 600 nm, consistent with Co2+ binding to the desired scaffolded motif (33). The mutant design was the same sequence but with the 6 coordinating residues (sidechains shown in (A)) mutated to alanine [E16A, E55A, H58A, E89A, H92A, E115A]). Protein concentration was 200 μM. (C) Titration analysis of Co2+ against the design (protein concentration = 200 μM). Quantification of the absorbance at 550 nm, using a predicted extinction coefficient of 155 for Co2+ binding the motif (33), is consistent with both binding sites being recapitulated in the dife_inp_1 design. (D) CD spectra of design in the presence and absence of Co2+. Both spectra are consistent with the predicted helical structure. (E) CD melt curve in the presence and absence of Co2+. Note that the coordination of Co2+ in the protein core significantly stabilizes dife_inp_1 (protein concentration in CD experiments = 6.7 μM, Co2+ concentration = 53.3 μM). (F) AF2 prediction of inpainted design EFhand_inp_1 scaffolding the double EF-hand motif with input motif residues in purple, input non-motif residues in green, and overlaid with the native motif from 1PRW (orange). (G) Tryptophan-enhanced terbium fluorescence spectra of EFhand_inp_1 matches known spectra (57) and suggests the design can bind terbium. (H) CD spectra of EFhand_inp_1 incubated with (4X protein concentration) and without CaCl2 suggest stabilization of the protein upon binding calcium. Design metrics (AF pLDDT, motif RMSD AF versus native): dife_inp_1 (92 /0.65 Å), EFhand_inp1 (84, 0.7 Å).](https://figures.cochat.ai/paper-figures/eb1c891ce6cb4e9a80924606dd10d2aa/scaffolding-protein-functional-sites-using-deep-learning-figure-3.jpg)


Verification
Claims the paper makes
9 Valid · 0 Not valid · 12 Not checked · 71 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“trRosetta can also be used to design sequences that fold into a target backbone structure by carrying out sequence optimization using a structure recapitulation loss function that rewards similarity of the predicted structure to the target structure (13).”
§ Scaffolding protein functional sites using deep learning > Abstract
What the sources say
- Supportedcites [13] Norn C, Wicky BIM, Juergens D, Liu S, Kim D, Tischer D, Koepnick B, Anishchenko I, Players F, Baker D, Ovchinnikov S, Protein sequence design by conformational landscape optimization. Proc. Natl. Acad. Sci. 118 (2021), doi: 10.1073/pnas.2017228118. PMC7980421 33712545 source
The Results and Discussions section, “Sequence Design by Gradient Backpropagation” (page not provided), describes optimizing amino acid sequences with a loss function based on the probabilities of the target structure’s observed interresidue distances and orientations. This supports the claim’s description of a structure-recapitulation loss that rewards agreement with the target backbone. (LLM, conf=0.91)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“We previously demonstrated that the trRosetta structure-prediction neural network (11) can be used to generate new proteins by maximizing the trRosetta output probability that a sequence folds to some (unspecified) three dimensional structure during Monte Carlo sampling in sequence space (12).”
§ Scaffolding protein functional sites using deep learning > Abstract
What the sources say
- Not checked yetcites [11] Yang J, Anishchenko I, Park H, Peng Z, Ovchinnikov S, Baker D, Improved protein structure prediction using predicted interresidue orientations. Proc. Natl. Acad. Sci. (2020), doi: 10.1073/pnas.1914677117. PMC6983395 31896580 source
- Partly supportedcites [12] Anishchenko I, Pellock SJ, Chidyausiku TM, Ramelot TA, Ovchinnikov S, Hao J, Bafna K, Norn C, Kang A, Bera AK, DiMaio F, Carter L, Chow CM, Montelione GT, Baker D, De novo protein design by deep network hallucination. Nature. 600, 547– 552 (2021). 34853475 PMC9293396 sourceExperimental — accuracy not yet measured
Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Previous methods can scaffold functional sites made up of one or two contiguous chain segments (6–10), but with the exception of helical bundles (8) these do not extend readily to more complex sites composed of three or more chain segments, and the generated backbones are not guaranteed to be designable (encodable by some amino acid sequence).”
§ Scaffolding protein functional sites using deep learning > Abstract
What the sources say
- Supportedcites [6] Procko E, Berguig GY, Shen BW, Song Y, Frayo S, Convertine AJ, Margineantu D, Booth G, Correia BE, Cheng Y, Schief WR, Hockenbery DM, Press OW, Stoddard BL, Stayton PS, Baker D, A Computationally Designed Inhibitor of an Epstein-Barr Viral Bcl-2 Protein Induces Apoptosis in Infected Cells. Cell. 157, 1644– 1656 (2014). 24949974 PMC4079535 source
The citation to this paper is attached to the helical-bundle exception, rather than to the general claim about the limits of previous methods. In the Results sections on grafting and de novo design (page not provided), the paper describes using 3- and 4-helix bundles to build a binding interface with contacts beyond the grafted BH3 helix. (LLM, conf=0.86)
- Could not verifycites [7] Correia BE, Bates JT, Loomis RJ, Baneyx G, Carrico C, Jardine JG, Rupert P, Correnti C, Kalyuzhniy O, Vittal V, Connell MJ, Stevens E, Schroeter A, Chen M, MacPherson S, Serra AM, Adachi Y, Holmes MA, Li Y, Klevit RE, Graham BS, Wyatt RT, Baker D, Strong RK, Crowe JE, Johnson PR, Schief WR, Proof of principle for epitope-focused vaccine design. Nature. 507, 201– 206 (2014). 24499818 PMC4260937 sourceExperimental — accuracy not yet measured
The Design of Epitope-Scaffolds section describes a helix-turn-helix epitope and a three-helix-bundle template, while Computational Method describes the paper’s Fold from Loops approach. Neither passage assesses whether previous methods extend to functional sites made of three or more chain segments; searches for that limit and the helical-bundle exception found no supporting passage in our copy. (LLM, conf=0.72)
- Supportedcites [8] Silva D-A, Yu S, Ulge UY, Spangler JB, Jude KM, Labão-Almeida C, Ali LR, Quijano-Rubio A, Ruterbusch M, Leung I, Biary T, Crowley SJ, Marcos E, Walkey CD, Weitzner BD, Pardo-Avila F, Castellanos J, Carter L, Stewart L, Riddell SR, Pepper M, Bernardes GJL, Dougan M, Garcia KC, Baker D, De novo design of potent and selective mimics of IL-2 and IL-15. Nature. 565, 186– 191 (2019). 30626941 PMC6521699 source
The citation is attached to the helical-bundle exception. In “Computational design of IL-2/IL-15 mimics that bind and activate IL-2Rβγc” (page unavailable), the paper describes building a connected protein around the three helices H1, H3 and H4 that form the receptor-binding interface, supporting that exception. (LLM, conf=0.91)
- Partly supportedcites [9] Sesterhenn F, Yang C, Bonet J, Cramer JT, Wen X, Wang Y, Chiang C-I, Abriata LA, Kucharska I, Castoro G, Vollers SS, Galloux M, Dheilly E, Rosset S, Corthésy P, Georgeon S, Villard M, Richard C-A, Descamps D, Delgado T, Oricchio E, Rameix-Welti M-A, Más V, Ervin S, Eléouët J-F, Riffault S, Bates JT, Julien J-P, Li Y, Jardetzky T, Krey T, Correia BE, De novo protein design enables the precise induction of RSV-neutralizing antibodies. Science. 368 (2020), doi: 10.1126/science.aay5051. PMC7391827 32409444 sourceExperimental — accuracy not yet measured
The Introduction (page not provided) says earlier grafting work largely used simple motifs such as linear helical segments and identifies irregular, multi-segment functional sites as a challenge. The Discussion & Conclusions (page not provided) also describes limitations of template-based design. Neither passage establishes the specific three-or-more-segment threshold or identifies helical bundles as an exception. (LLM, conf=0.82)
- Not checked yetcites [10] Yang C, Sesterhenn F, Bonet J, van Aalen EA, Scheller L, Abriata LA, Cramer JT, Wen X, Rosset S, Georgeon S, Jardetzky T, Krey T, Fussenegger M, Merkx M, Correia BE, Bottom-up de novo design of functional proteins with complex structural features. Nat. Chem. Biol, 1– 9 (2021). 33398169 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Previous methods can scaffold functional sites made up of one or two contiguous chain segments (6–10), but with the exception of helical bundles (8) these do not extend readily to more complex sites composed of three or more chain segments, and the generated backbones are not guaranteed to be designable (encodable by some amino acid sequence).”
§ Scaffolding protein functional sites using deep learning > Abstract
What the sources say
- Partly supportedcites [6] Procko E, Berguig GY, Shen BW, Song Y, Frayo S, Convertine AJ, Margineantu D, Booth G, Correia BE, Cheng Y, Schief WR, Hockenbery DM, Press OW, Stoddard BL, Stayton PS, Baker D, A Computationally Designed Inhibitor of an Epstein-Barr Viral Bcl-2 Protein Induces Apoptosis in Infected Cells. Cell. 157, 1644– 1656 (2014). 24949974 PMC4079535 sourceExperimental — accuracy not yet measured
The Results section, under “BHRF1-binding proteins created by grafting Bim-BH3 side chains to a helical scaffold” (page unavailable), describes grafting the interacting residues of a single BH3 helix onto protein scaffolds. This supports the one-contiguous-segment example, but our copy does not establish that previous methods scaffold functional sites made up of two contiguous chain segments. (LLM, conf=0.82)
- Supportedcites [7] Correia BE, Bates JT, Loomis RJ, Baneyx G, Carrico C, Jardine JG, Rupert P, Correnti C, Kalyuzhniy O, Vittal V, Connell MJ, Stevens E, Schroeter A, Chen M, MacPherson S, Serra AM, Adachi Y, Holmes MA, Li Y, Klevit RE, Graham BS, Wyatt RT, Baker D, Strong RK, Crowe JE, Johnson PR, Schief WR, Proof of principle for epitope-focused vaccine design. Nature. 507, 201– 206 (2014). 24499818 PMC4260937 source
The Abstract (page not available) says that previously developed side-chain and backbone grafting methods transplanted continuous or discontinuous epitopes onto scaffold proteins. This supports the attributed point that previous methods could scaffold functional sites comprising contiguous chain segments; the claim does not attribute a limitation to exactly two segments to this paper. (LLM, conf=0.86)
- Could not verifycites [8] Silva D-A, Yu S, Ulge UY, Spangler JB, Jude KM, Labão-Almeida C, Ali LR, Quijano-Rubio A, Ruterbusch M, Leung I, Biary T, Crowley SJ, Marcos E, Walkey CD, Weitzner BD, Pardo-Avila F, Castellanos J, Carter L, Stewart L, Riddell SR, Pepper M, Bernardes GJL, Dougan M, Garcia KC, Baker D, De novo design of potent and selective mimics of IL-2 and IL-15. Nature. 565, 186– 191 (2019). 30626941 PMC6521699 sourceExperimental — accuracy not yet measured
The computational-design section describes this paper’s method for scaffolding an IL-2 binding interface formed by three helices (H1, H3 and H4). It does not address whether previous methods could scaffold functional sites made up of one or two contiguous chain segments; searches for prior scaffolding methods found no passage establishing that point. The paper does, however, fit the helical-bundle exception separately identified in the full citing sentence. (LLM, conf=0.76)
- Supportedcites [9] Sesterhenn F, Yang C, Bonet J, Cramer JT, Wen X, Wang Y, Chiang C-I, Abriata LA, Kucharska I, Castoro G, Vollers SS, Galloux M, Dheilly E, Rosset S, Corthésy P, Georgeon S, Villard M, Richard C-A, Descamps D, Delgado T, Oricchio E, Rameix-Welti M-A, Más V, Ervin S, Eléouët J-F, Riffault S, Bates JT, Julien J-P, Li Y, Jardetzky T, Krey T, Correia BE, De novo protein design enables the precise induction of RSV-neutralizing antibodies. Science. 368 (2020), doi: 10.1126/science.aay5051. PMC7391827 32409444 source
The Introduction (page unavailable) describes earlier grafting methods using linear helical segments. The Results section, ‘De novo design of immunogens presenting structurally complex epitopes’ (page unavailable), adds that one earlier study grafted a discontinuous HIV epitope into a scaffold. Together, these support the claim that previous methods could scaffold sites comprising a single segment or more than one segment. (LLM, conf=0.85)
- Not checked yetcites [10] Yang C, Sesterhenn F, Bonet J, van Aalen EA, Scheller L, Abriata LA, Cramer JT, Wen X, Rosset S, Georgeon S, Jardetzky T, Krey T, Fussenegger M, Merkx M, Correia BE, Bottom-up de novo design of functional proteins with complex structural features. Nat. Chem. Biol, 1– 9 (2021). 33398169 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Previous methods can scaffold functional sites made up of one or two contiguous chain segments (6–10), but with the exception of helical bundles (8) these do not extend readily to more complex sites composed of three or more chain segments, and the generated backbones are not guaranteed to be designable (encodable by some amino acid sequence).”
§ Scaffolding protein functional sites using deep learning > Abstract
What the sources say
- Partly supportedcites [6] Procko E, Berguig GY, Shen BW, Song Y, Frayo S, Convertine AJ, Margineantu D, Booth G, Correia BE, Cheng Y, Schief WR, Hockenbery DM, Press OW, Stoddard BL, Stayton PS, Baker D, A Computationally Designed Inhibitor of an Epstein-Barr Viral Bcl-2 Protein Induces Apoptosis in Infected Cells. Cell. 157, 1644– 1656 (2014). 24949974 PMC4079535 sourceExperimental — accuracy not yet measured
The Results section, under “In silico folding probability correlates with binding activity” (page unavailable), says that functional protein design has a low success rate and that inactive designs may fold improperly or have unstable designed states. This supports the general point that a generated design is not guaranteed to work, but it does not establish the more specific claim that a generated backbone cannot be encoded by any amino acid sequence. (LLM, conf=0.79)
- Could not verifycites [7] Correia BE, Bates JT, Loomis RJ, Baneyx G, Carrico C, Jardine JG, Rupert P, Correnti C, Kalyuzhniy O, Vittal V, Connell MJ, Stevens E, Schroeter A, Chen M, MacPherson S, Serra AM, Adachi Y, Holmes MA, Li Y, Klevit RE, Graham BS, Wyatt RT, Baker D, Strong RK, Crowe JE, Johnson PR, Schief WR, Proof of principle for epitope-focused vaccine design. Nature. 507, 201– 206 (2014). 24499818 PMC4260937 sourceExperimental — accuracy not yet measured
The Abstract (page unavailable) says the earlier grafting methods used scaffold proteins of known structure, but does not say whether their generated backbones were guaranteed to be designable. The Computational Method and Methods > Fold From Loops sections describe sequence design and filtering for the new method; a search for designability of backbones from previous methods found no passage addressing that point. (LLM, conf=0.72)
- Could not verifycites [8] Silva D-A, Yu S, Ulge UY, Spangler JB, Jude KM, Labão-Almeida C, Ali LR, Quijano-Rubio A, Ruterbusch M, Leung I, Biary T, Crowley SJ, Marcos E, Walkey CD, Weitzner BD, Pardo-Avila F, Castellanos J, Carter L, Stewart L, Riddell SR, Pepper M, Bernardes GJL, Dougan M, Garcia KC, Baker D, De novo design of potent and selective mimics of IL-2 and IL-15. Nature. 565, 186– 191 (2019). 30626941 PMC6521699 sourceExperimental — accuracy not yet measured
The Methods section (page unavailable) describes how this paper generated backbones and then performed sequence design and filtering. Searches for prior scaffolding methods and backbone designability found no passage saying that backbones generated by previous methods are not guaranteed to be designable. (LLM, conf=0.72)
- Could not verifycites [9] Sesterhenn F, Yang C, Bonet J, Cramer JT, Wen X, Wang Y, Chiang C-I, Abriata LA, Kucharska I, Castoro G, Vollers SS, Galloux M, Dheilly E, Rosset S, Corthésy P, Georgeon S, Villard M, Richard C-A, Descamps D, Delgado T, Oricchio E, Rameix-Welti M-A, Más V, Ervin S, Eléouët J-F, Riffault S, Bates JT, Julien J-P, Li Y, Jardetzky T, Krey T, Correia BE, De novo protein design enables the precise induction of RSV-neutralizing antibodies. Science. 368 (2020), doi: 10.1126/science.aay5051. PMC7391827 32409444 sourceExperimental — accuracy not yet measured
The Introduction and Discussion & Conclusions describe limitations of earlier scaffolding approaches, including destabilization and the need for extensive optimization. Neither passage says whether the generated backbones are guaranteed to be encodable by an amino acid sequence; a search for backbone designability found no such statement in our copy. (LLM, conf=0.72)
- Not checked yetcites [10] Yang C, Sesterhenn F, Bonet J, van Aalen EA, Scheller L, Abriata LA, Cramer JT, Wen X, Rosset S, Georgeon S, Jardetzky T, Krey T, Fussenegger M, Merkx M, Correia BE, Bottom-up de novo design of functional proteins with complex structural features. Nat. Chem. Biol, 1– 9 (2021). 33398169 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
59 more claims · 3 with their cited sources checked
Create a free account to see all 92 claims and to verify the 16 references still open.
- Probably not a substantial claim. 5
- It may state two separate claims as one. 23
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Experimental characterization showed that the method can scaffold a wide range of protein functional motifs with atomic accuracy19”
From De novo design of protein structure and function with RFdiffusion (2023) · cites this paper as [4]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“but the approach fails on minimalist site descriptions that do not sufficiently constrain the overall fold and, because it is deterministic, can produce only a limited diversity of designs for a given problem.”
From De novo design of protein structure and function with RFdiffusion (2023) · cites this paper as [4]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Experimental characterization showed that the method can scaffold a wide range of protein functional motifs with atomic accuracy19”
From De novo design of protein structure and function with RFdiffusion (2023) · cites this paper as [19]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“This measure of in silico success has been found to correlate with experimental success4,7,26 and is significantly more stringent than template modelling (TM)-score-based metrics used elsewhere5,16,27–29”
From De novo design of protein structure and function with RFdiffusion (2023) · cites this paper as [4]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“we established an in silico benchmark test (Supplementary Table 9) comprising 25 motif-scaffolding design problems addressed in six recent publications encompassing several design methodologies4,5,29,38–40”
From De novo design of protein structure and function with RFdiffusion (2023) · cites this paper as [4]
Experimental — accuracy not yet measured
Citations
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Basic local alignment search tool
Altschul, S. F., Gish, W., Miller, W., Myers, E. W., & Lipman, D. J. (1990) · Journal of Molecular Biology · doi:10.1006/jmbi.1990.9999 · Cited by 98,669
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
Highly accurate protein structure prediction with AlphaFold
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03819-2 · Cited by 43,764
Berman, H. M. (2000) · Nucleic Acids Research · doi:10.1093/nar/28.1.235 · Cited by 40,196
Comparison of multiple Amber force fields and development of improved protein backbone parameters
Hornak, V., Abel, R., Okur, A., Strockbine, B., Roitberg, A., & Simmerling, C. (2006) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.21123 · Cited by 6,914
Protein production by auto-induction in high-density shaking cultures
Studier, F. W. (2005) · Protein Expression and Purification · doi:10.1016/j.pep.2005.01.016 · Cited by 5,931
Accurate prediction of protein structures and interactions using a 3-track network
Baek, M., DiMaio, F., Anishchenko, I., Dauparas, J., Ovchinnikov, S., Lee, G. R., Wang, J., Cong, Q., Kinch, L. N., Schaeffer, R. D., et al. (2021) · doi:10.1101/2021.06.14.448402 · Cited by 5,549 · cited 3×
“Following the development of RosettaFold (RF) (16) we found that it performed better than trRosetta in guiding protein design by functional-site-constrained hallucination (Fig.”
BackgroundMethodInfluentialSelf-citationMMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets
Steinegger, M., & Söding, J. (2017) · Nature Biotechnology · doi:10.1038/nbt.3988 · Cited by 4,840
Research works printed but not linked: 1
- Z. Lin T. Sercu Y. LeCun A. Rives “Deep generative models create new and diverse protein structures ” 35th Conference on Neural Information Processing Systems (NeurIPS 2021) 6 to 14 December 2021.
Cited by
Items 101–200 of 492
Accurate prediction of protein function using statistics-informed graph networks
Jang, Y. J., Qin, Q.-Q., Huang, S., Peter, A. T. J., Ding, X., & Kornmann, B. (2024) · Nature Communications · doi:10.1038/s41467-024-50955-0 · Cited by 60
“This evolutionary data, especially the couplings between pairwise residues, has also been utilized to characterize protein functional sites 10,11 .”
Generative artificial intelligence for enzyme design: Recent advances in models and applications
Wen, S., Zheng, W., Bornscheuer, U. T., & Wu, S. (2025) · Current Opinion in Green and Sustainable Chemistry · doi:10.1016/j.cogsc.2025.101010 · Cited by 61
An end-to-end framework for the prediction of protein structure and fitness from single sequence
Chen, Y., Xu, Y., Liu, D., Xing, Y., & Gong, H. (2024) · Nature Communications · doi:10.1038/s41467-024-51776-x · Cited by 45
“…SPIRED-Fitness using the single mutational data in the training set and evaluated its performance against unsupervised models including RF joint 17,41 , MSA Transformer 6 , ESM-2 11 , ESM-1b 5 , ESM-1v 42 , and DeepSequence 40 as well as supervised models including ECNet 36 and SESNet 19 also on…”
Zero‐shot mutation effect prediction on protein stability and function using RoseTTAFold
Mansoor, S., Baek, M., Juergens, D., Watson, J. L., & Baker, D. (2023) · Protein Science · doi:10.1002/pro.4780 · Cited by 40
Self-citationGao, T., Huang, H., & Liu, Y. (2025) · Advanced Materials · doi:10.1002/adma.202508263 · Cited by 52
Wang, J., Watson, J. L., & Lisanza, S. L. (2024) · Cold Spring Harbor Perspectives in Biology · doi:10.1101/cshperspect.a041472 · Cited by 48
Self-citationRapid and automated design of two-component protein nanomaterials using ProteinMPNN
de Haas, R. J., Brunette, N., Goodson, A., Dauparas, J., Yi, S. Y., Yang, E. C., Dowling, Q., Nguyen, H., Kang, A., Bera, A. K., et al. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2314646121 · Cited by 42
Self-citationTowards rational computational peptide design
Chang, L., Mondal, A., & Pérez, A. (2022) · Frontiers in Bioinformatics · doi:10.3389/fbinf.2022.1046493 · Cited by 48
“Recent protein design work is already leveraging machine learning to hallucinate novel folds based on the idealized version of proteins learnt by the algorithms (Anishchenko et al., 2021; Wang et al., 2022).”
BackgroundProtein Language Models: Is Scaling Necessary?
Fournier, Q., Vernon, R. M., van der Sloot, A., Schulz, B., Chandar, S., & Langmead, C. J. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.09.23.614603 · Cited by 26
“…in using machine learning approaches to address protein structure prediction and design have emerged as a direct consequence of this shift (Senior et al., 2020; Jumper et al., 2021; Dauparas et al., 2022; Wang et al., 2022; Ingraham et al., 2023; Watson et al., 2023; Abramson et al., 2024).”
Rapid and automated design of two-component protein nanomaterials using ProteinMPNN
de Haas, R. J., Brunette, N., Goodson, A., Dauparas, J., Yi, S. Y., Yang, E. C., Dowling, Q., Nguyen, H., Kang, A., Bera, A. K., et al. (2023) · doi:10.1101/2023.08.04.551935 · Cited by 42
Self-citationXu, Y., Liu, D., & Gong, H. (2024) · Nature Computational Science · doi:10.1038/s43588-024-00716-2 · Cited by 41
“Another strategy is to train the model with the masked language modeling objectives, like in RF joint 22,23 , MSA transformer 18 , ESM-1b 24 , ESM-1v 17 and ESM-2 25 , which completely bypasses the need of any fitness data.”
BackgroundLin, M., Yin, Y., Zhao, X., Wang, C., Zhu, X., Zhan, L., Chen, L., Wang, S., Lin, X. M., Zhang, J., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-56302-1 · Cited by 46
“The RSV pre-F trimer consists of a “ stalk ” and a “ head ” region, with antigenic sites Φ and V located on the head, recognized by potent nAbs such as D25 12 and hRSV90 23,24 .”
Malbranke, C., Bikard, D., Cocco, S., Monasson, R., & Tubiana, J. (2023) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2023.102571 · Cited by 35
Peptide binder design with inverse folding and protein structure prediction
Bryant, P., & Elofsson, A. (2023) · Communications Chemistry · doi:10.1038/s42004-023-01029-7 · Cited by 49
“These methods have been applied to design binders with and without scaffolds and known binding motifs 15 .”
Design of high-affinity binders to immune modulating receptors for cancer immunotherapy
Yang, W., Hicks, D. R., Ghosh, A., Schwartze, T. A., Conventry, B., Goreshnik, I., Allen, A., Halabiya, S. F., Kim, C. J., Hinck, C. S., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-57192-z · Cited by 40
Self-citationYadav, S. B., Lachowski, K. J., Chiang, H. T., Pozzo, L. D., Yoreo, J. D., & Zhang, S. (2023) · Angewandte Chemie International Edition · doi:10.1002/anie.202309725 · Cited by 39
Cytochrome P450 Enzyme Design by Constraining the Catalytic Pocket in a Diffusion Model
Wang, Q., Liu, X., Zhang, H., Chu, H., Shi, C., Zhang, L. J., Liu, P., Liu, P., Zhu, X., Liu, Y., et al. (2024) · Research · doi:10.34133/research.0413 · Cited by 43
“The rapid development of deep learning has witnessed many impressive achievements in protein structure design, while the desired functional design is still a big challenge [55–57].”
BackgroundAgajanian, S., Alshahrani, M., Bai, F., Tao, P., & Verkhivker, G. M. (2023) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.2c01634 · Cited by 41
AlphaDesign: a de novo protein design framework based on AlphaFold
(2025) · doi:10.1038/s44320-025-00119-z · Cited by 36
“Currently, only the structure of RcaT-Eco1 has been experimentally determined (Wang et al, 2022; Carabias et al, 2024).”
Akpinaroglu, D., Seki, K., Guo, A., Zhu, E., Kelly, M. J. S., & Kortemme, T. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.12.15.571823 · Cited by 33
The role and pathway of VQ family in plant growth, immunity, and stress response
Tian, J., Zhang, J., & Francis, F. (2023) · Planta · doi:10.1007/s00425-023-04292-z · Cited by 34
“…applied in protein structure prediction and new functional protein development (Baek et al. 2021; Tunyasuvunakool et al. 2021; Bryant et al. 2022; Dauparas et al. 2022; Wang et al. 2022b; Huang et al. 2023; Lutz et al. 2023; Madani et al. 2023; Schuster et al. 2023; Watson et al. 2023).”
BackgroundStructural modeling of ion channels using AlphaFold2, RoseTTAFold2, and ESMFold
Nguyen, P. T., Harris, B., Mateos, D. L., González, A. H., Murray, A., & Yarov-Yarovoy, V. M. (2024) · Channels · doi:10.1080/19336950.2024.2325032 · Cited by 40
Are fibrinaloid microclots a cause of autoimmunity in Long Covid and other post-infection diseases?
Kell, D. B., & Pretorius, E. (2023) · Biochemical Journal · doi:10.1042/bcj20230241 · Cited by 34
Chao, Y., Han, Y. J., Chen, Z., Chu, D., Xu, Q. J., Wallace, G. G., & Wang, C. (2023) · Advanced Science · doi:10.1002/advs.202305558 · Cited by 36
EvoBind: in silico directed evolution of peptide binders with AlphaFold
Bryant, P., & Elofsson, A. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.07.23.501214 · Cited by 29
“These methods have been applied to design binders both with and without scaffolds and known binding motifs[15].”
MethodXu, X., Xu, T., Zhou, J., Liao, X., Zhang, R., Wang, Y., Zhang, L., & Gao, X. (2023) · Genomics, Proteomics & Bioinformatics · doi:10.1016/j.gpb.2023.03.004 · Cited by 32
“For example, RFDesign was proposed to design proteins with specific functions, such as immunogen, enzyme activity, and protein–protein interaction [10].”
BackgroundMethodEnzyme Databases in the Era of Omics and Artificial Intelligence
Prešern, U., & Goličnik, M. (2023) · International Journal of Molecular Sciences · doi:10.3390/ijms242316918 · Cited by 33
“Deep learning algorithms for the prediction of enzyme function [46–49], structure [50,51], catalytic residues [52,53], and kinetic properties [54–56], as well as algorithms for the de novo design of enzymes [57,58], have already been developed.”
BackgroundDesign of facilitated dissociation enables timing of cytokine signalling
Broerman, A. J., Pollmann, C., Zhao, Y., Lichtenstein, M. A., Jackson, M. D., Tessmer, M. H., Ryu, W. H., Ogishi, M., Abedi, M. H., Sahtoe, D. D., et al. (2025) · Nature · doi:10.1038/s41586-025-09549-z · Cited by 37
“We then used inpainting with RosettaFold to generate loops connecting the domains, then paired these complete state X backbones with the original effector-bound state Y of AS0 (also generating new loops between domains in state Y as necessary to match the loop lengths of each new state X).”
Self-citationBeyond sequence: Structure-based machine learning
Durairaj, J., de Ridder, D., & van Dijk, A. D. J. (2023) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2022.12.039 · Cited by 33
Generative models for protein structures and sequences
Hsu, C., Fannjiang, C., & Listgarten, J. (2024) · Nature Biotechnology · doi:10.1038/s41587-023-02115-w · Cited by 38
Computational Design of Peptide Assemblies
Min, J., Rong, X., Zhang, J., Su, R., Wang, Y., & Qi, W. D. (2024) · Journal of Chemical Theory and Computation · doi:10.1021/acs.jctc.3c01054 · Cited by 32
Protein‐Based Controllable Nanoarchitectonics for Desired Applications
Li, L., Zhang, Y., Wu, Y., Wang, Z., Cui, W., Zhang, C., Wang, J., Liu, Y., & Yang, P. (2024) · Advanced Functional Materials · doi:10.1002/adfm.202315509 · Cited by 37
Self-citationZahid, A., Ismail, H., Wilson, J. C., & Grice, I. D. (2025) · Vaccines · doi:10.3390/vaccines13070767 · Cited by 38
Artificial Intelligence-Driven Enzyme Engineering from Structural Prediction to De Novo Design
Shi, H., Bai, X., Tian, F., Li, Y., Li, D., Yao, L., Xue, C., & Tang, C. (2026) · Journal of Agricultural and Food Chemistry · doi:10.1021/acs.jafc.6c01781 · Cited by 32
Dou, Z., Sun, Y., Jiang, X., Wu, X., Li, Y., Gong, B., & Wang, L. (2023) · Acta Biochimica et Biophysica Sinica · doi:10.3724/abbs.2023033 · Cited by 34
Rationally seeded computational protein design of ɑ-helical barrels
Albanese, K. I., Petrenas, R., Pirro, F., Naudin, E. A., Borucu, U., Dawson, W., Scott, D. A., Leggett, G. J., Weiner, O. D., Oliver, T. A. A., et al. (2024) · Nature Chemical Biology · doi:10.1038/s41589-024-01642-0 · Cited by 33
“9, 34 and 35).”
From traditional to AI-driven: The evolution of intelligent enzyme engineering for biocatalysis
Lv, F., Zhang, J., You, S., & Qi, W. (2025) · Biotechnology Advances · doi:10.1016/j.biotechadv.2025.108788 · Cited by 36
Protein Function Analysis through Machine Learning
Avery, C., Patterson, J., Grear, T. J., Frater, T. D., & Jacobs, D. J. (2022) · Biomolecules · doi:10.3390/biom12091246 · Cited by 31
A De Novo‐Designed Type 3 Copper Protein Tunes Catechol Substrate Recognition and Reactivity
Pirro, F., Gatta, S. L., Arrigoni, F., Famulari, A., Maglio, O., Vecchio, P. D., Chiesa, M., GIOIA, L. D., Bertini, L., Chino, M., et al. (2022) · Angewandte Chemie International Edition · doi:10.1002/anie.202211552 · Cited by 30
Chen, Y., Zhang, S., Zhai, Z., Zhang, S., Ma, J., Liang, X., & Li, Q. (2023) · International Journal of Molecular Sciences · doi:10.3390/ijms24032780 · Cited by 36
“The additional structure could stabilize the catalytic domain of fusion proteins, thereby improving the thermal stability of enzymes [39,40].”
BackgroundAI-assisted food enzymes design and engineering: a critical review
Wang, X., Yang, P., Zhao, B., & Liu, S. (2022) · Systems Microbiology and Biomanufacturing · doi:10.1007/s43393-022-00138-z · Cited by 30
“Another breakthrough in enzyme design is de novo design sequences based on an input sequence, the AI tools, such as “Hallucination” and proteinGAN, have enabled such functions [21, 95, 96].”
BackgroundCell-cultivated aquatic food products: emerging production systems for seafood
Goswami, M., Ovissipour, R., Bomkamp, C., Nitin, N., Lakra, W. S., Post, M. J., & Kaplan, D. L. (2024) · Journal of Biological Engineering · doi:10.1186/s13036-024-00436-1 · Cited by 31
“Approaches to scaffolding and tissue engineering for cell-cultivated meat have been reviewed elsewhere [54–57].”
BackgroundArtificial Intelligence-Driven de Novo Design of Robust Enzymes to Enhance Their Performance
Tang, M., Ge, F., Li, A., Hu, L., Wang, C., Tang, J., Song, X., Liu, X., Shi, H. S., & Tan, Z. (2025) · ACS Synthetic Biology · doi:10.1021/acssynbio.5c00452 · Cited by 31
Deep learning for metabolic pathway design
Ryu, G. U., Kim, G. B., Yu, T., & Lee, S. Y. (2023) · Metabolic Engineering · doi:10.1016/j.ymben.2023.09.012 · Cited by 30
Diagnosis, Treatment, and Management of Otitis Media with Artificial Intelligence
Ding, X., Huang, Y., Tian, X., Zhao, Y. S., Feng, G., & Gao, Z. (2023) · Diagnostics · doi:10.3390/diagnostics13132309 · Cited by 33
“, devised RoseTTAFold, a “three-track” neural network capable of analysing the three-dimensional structure of a sequence of proteins within 20 min [76].”
MethodDirected evolution of a highly efficient TNA polymerase achieved by homologous recombination
Maola, V. A., Yik, E. J., Hajjar, M., Lee, J. J., Holguin, M. J., Quijano, R. N., Nguyen, K. K., Ho, K. L., Medina, J. V., Chim, N., et al. (2024) · Nature Catalysis · doi:10.1038/s41929-024-01233-1 · Cited by 30
“Although computational methods have solved many problems in protein design 4,5 , attempts to create XNA polymerases by in silico prediction have thus far proven unsuccessful, presumably due to the complicated reaction pathway 6 and dynamic motion 7 that replicative DNA polymerases follow for each…”
Enhancement Strategies of Calcium Looping Technology and CaO‐Based Sorbents for Carbon Capture
Wang, Z., Ma, C., Harrison, A. R. P., Alsouleman, K., Gao, M., Huang, Z., Chen, Q., & Nie, B. (2025) · Small · doi:10.1002/smll.202412463 · Cited by 42
Mahling, R. W., Hegyi, B., Cullen, E. R., Cho, T. M., Rodriques, A., Fossier, L., Yehya, M., Yang, L., Chen, B.-X., Katchman, A. N., et al. (2025) · Cell · doi:10.1016/j.cell.2025.07.038 · Cited by 29
Self-citationAtomically Accurate Design of Metalloproteins with Predefined Coordination Geometries
Hoffnagle, A. M., & Tezcan, F. A. (2023) · Journal of the American Chemical Society · doi:10.1021/jacs.3c04047 · Cited by 27
IgGM: A Generative Model for Functional Antibody and Nanobody DesignPreprint
Wang, R., Wu, F., Gao, X., Wu, J., Zhao, P., & Yao, J. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.09.19.613838 · Cited by 29
Garcia, M., Dixit, S. M., & Rocklin, G. J. (2026) · Protein Science · doi:10.1002/pro.70453 · Cited by 36
“…language models such as ESMFold have been integrated in design strategies as final validation steps to remove poor designs from experimental characterization (Abramson et al. 2024; Dauparas et al. 2022; Jumper et al. 2021; Lin et al. 2023; Sumida et al. 2024; Wang et al. 2022; Watson et al. 2023).”
HelixGAN a deep-learning methodology for conditional de novo design of α-helix structures
Xie, X., Valiente, P. A., & Kim, P. M. (2023) · Bioinformatics · doi:10.1093/bioinformatics/btad036 · Cited by 32
Efficient and scalable de novo protein design using a relaxed sequence space
Frank, C., Khoshouei, A., de Stigter, Y., Schiewitz, D., Feng, S., Ovchinnikov, S., & Dietz, H. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.02.24.529906 · Cited by 26
Self-citationAn integrative approach to protein sequence design through multiobjective optimization
Hong, L., & Kortemme, T. (2024) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1011953 · Cited by 28
Accelerating protein design by scaling experimental characterization
Qian, J., Milles, L. F., Wicky, B. I. M., Ragotte, R. J., Motmaen, A., Borst, A. J., Skotheim, R., Ols, S., Coventry, B., Li, X., et al. (2026) · Nature Communications · doi:10.1038/s41467-026-76740-9 · Cited by 29
Self-citationFrom peptides to proteins: coiled-coil tetramers to single-chain 4-helix bundles
Naudin, E. A., Albanese, K. I., Smith, A. J., Mylemans, B., Baker, E. G., Weiner, O. D., Andrews, D. M., Tigue, N., Savery, N. J., & Woolfson, D. N. (2022) · Chemical Science · doi:10.1039/d2sc04479j · Cited by 24
Determinants for an Efficient Enzymatic Catalysis in Poly(Ethylene Terephthalate) Degradation
Castro-Rodríguez, J. A., Rodríguez‐Sotres, R., & Farrés, A. (2023) · Catalysts · doi:10.3390/catal13030591 · Cited by 25
Single missense mutations in Vi capsule synthesis genes confer hypervirulence to Salmonella Typhi
Lee, G. Y., & Song, J. (2024) · Nature Communications · doi:10.1038/s41467-024-49590-6 · Cited by 25
“Typhi, we predicted the 3-dimensional (D) structure of TviE by running two machine-learning-based structure prediction programs AlphaFold v2 and RosettaFold 55,56 (Fig.”
Protein-based materials: Applications, modification and molecular design
Tunuhe, A., Zheng, Z., Rao, X., Yu, H., Ma, F., Zhou, Y., & Xie, S. (2025) · BioDesign Research · doi:10.1016/j.bidere.2025.100004 · Cited by 24
“Deep learning is another powerful method that uses statistics to build models representing the real world through massive datasets, with neural networks being the primary tools [73,74].”
Müeller, U., Barthel, T., Benz, L. S., Bon, V., Crosskey, T., Dieguez, C. G., Förster, R., Gless, C., Hauß, T., Heinemann, U., et al. (2025) · Journal of Synchrotron Radiation · doi:10.1107/s1600577525001110 · Cited by 24
Deep Learning for Flexible and Site-Specific Protein Docking and Design
McPartlon, M., & Xu, J. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.04.01.535079 · Cited by 20
ESMFold Hallucinates Native-Like Protein SequencesPreprint
Jeliazkov, J. R., del Alamo, D., & Karpiak, J. D. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.05.23.541774 · Cited by 27
“AFDesign did not, consistent with previous reports [8, 12].”
MethodResultDesign of microbial catalysts for two-stage processes
Shabestary, K., Klamt, S., Link, H., Mahadevan, R., Steuer, R., & Hudson, E. P. (2024) · Nature Reviews Bioengineering · doi:10.1038/s44222-024-00225-x · Cited by 25
Of Revolutions and Roadblocks: The Emerging Role of Machine Learning in Biocatalysis
Vornholt, T., Stockinger, P., Mutný, M., Jeschek, M., Nestl, B. M., Oberdorfer, G., Osuna, S., Pleiss, J., Welner, D. H., Krause, A., et al. (2025) · ACS Central Science · doi:10.1021/acscentsci.5c00949 · Cited by 24
Small molecule allosteric modulation of the adenosine A1 receptor
Nguyen, A. T. N., Tran, Q. L., Baltos, J., McNeill, S. M., Nguyen, D. T. N., & May, L. T. (2023) · Frontiers in Endocrinology · doi:10.3389/fendo.2023.1184360 · Cited by 24
Self-assembly systems to troubleshoot metabolic engineering challenges
Chen, Z., Wu, T., Yu, S., Li, M., Fan, X., & Huo, Y. (2023) · Trends in biotechnology · doi:10.1016/j.tibtech.2023.06.009 · Cited by 24
Extracellular vesicles for developing targeted hearing loss therapy
Pan, X., Li, Y., Huang, P., Staecker, H., & He, M. (2024) · Journal of Controlled Release · doi:10.1016/j.jconrel.2023.12.050 · Cited by 25
Machine learning-aided design and screening of an emergent protein function in synthetic cells
Kohyama, S., Frohn, B. P., Babl, L., & Schwille, P. (2024) · Nature Communications · doi:10.1038/s41467-024-46203-0 · Cited by 23
“…the experimental screening of newly designed proteins is typically the expression of target proteins in E. coli cells followed by chromatographic puri fi cation protocols 6,7,25,27 this approach comes with many dif fi culties in experimental optimization due to protein solubility, cell toxicity, etc.”
BackgroundThe Role of AI-Driven De Novo Protein Design in the Exploration of the Protein Functional Universe
Zhang, G., Liu, C., Lu, J., Zhang, S., & Zhu, L. (2025) · Biology · doi:10.3390/biology14091268 · Cited by 24
“Mahling et al. used Colabdesign [147] to produce a 21-residue peptide (ELIXIR) that selectively inhibits pathological late or persistent Na + current ( I NaL ) by enhancing Na V 1.5 channel inactivation.”
InfluentialPossibilities of Using De Novo Design for Generating Diverse Functional Food Enzymes
Wang, X., Xu, K., Tan, Y., Liu, S., & Zhou, J. (2023) · International Journal of Molecular Sciences · doi:10.3390/ijms24043827 · Cited by 23
“In the past five years, achievements in using protein modeling and AI-based energy calculation for the rational design or de novo generation of functional proteins was conspicuous [22,33], especially for using de novo design to screen highaffinity protein binders, including drug proteins and antibodies [31,33].”
BackgroundMethodInfluentialComputational design of vaccine immunogens
Castro, K. M., Scheck, A., Xiao, S., & Correia, B. E. (2022) · Current Opinion in Biotechnology · doi:10.1016/j.copbio.2022.102821 · Cited by 25
Self-citationTAPIR: a T-cell receptor language model for predicting rare and novel targetsPreprint
Fast, E., Dhar, M., & Chen, B. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.09.12.557285 · Cited by 27
De novo design of high-affinity protein binders to bioactive helical peptides
Vázquez Torres, S., Leung, P. J. Y., Lutz, I. D., Venkatesh, P., Watson, J. L., Hink, F., Huynh, H.-H., Yeh, A. H.-W., Juergens, D., Bennett, N. R., et al. (2022) · doi:10.1101/2022.12.10.519862 · Cited by 22
“Deep learning methods do not have these limitations, and we explored whether RoseTTAFold inpainting (RF joint ) 14 , a model that can jointly design protein sequences and structures, could be used to improve the modest affinities of our parametrically-designed PTH binders (Fig.”
BackgroundMethodSelf-citationParametrically guided design of beta barrels and transmembrane nanopores using deep learning
Kim, D. E., Watson, J. L., Juergens, D., Majumder, S., Sonigra, R., Gerben, S. R., Kang, A., Bera, A. K., Li, X., & Baker, D. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2425459122 · Cited by 21
“In parallel, deep - learning methods have revolutionized de novo protein design, with methods such as Hallucination (13–17), RFjoint inpainting (14), and more recently generative models such as RFdiffusion (18) demonstrating improved performance across a range of protein design tasks as compared to…”
InfluentialSelf-citationProtein engineering of amine transaminases
Meng, Q., Ramírez-Palacios, C., Wijma, H. J., & Janssen, D. B. (2022) · Frontiers in Catalysis · doi:10.3389/fctls.2022.1049179 · Cited by 23
De Novo Design of Polyhedral Protein Assemblies: Before and After the AI Revolution
Mallik, B. B., Stanislaw, J., Alawathurage, T. M., & Khmelinskaia, A. (2023) · ChemBioChem · doi:10.1002/cbic.202300117 · Cited by 23
Fannjiang, C., & Listgarten, J. (2023) · Cold Spring Harbor Perspectives in Biology · doi:10.1101/cshperspect.a041469 · Cited by 25
Hallucinating structure-conditioned antibody libraries for target-specific binders
Mahajan, S. P., Ruffolo, J. A., Frick, R., & Gray, J. J. (2022) · Frontiers in Immunology · doi:10.3389/fimmu.2022.999034 · Cited by 20
“This idea of inverting neural networks has been previously tested with success with sequence-to-structure protein DL models to design de novo sequence-structure pairs (12), sequences that maximize the likelihood of a target structure (26), protein scaffolds that can host functional motifs (27).”
MethodFast, accurate ranking of engineered proteins by target-binding propensity using structure modeling
Ding, X., Chen, X., Sullivan, E. E., Shay, T. F., & Gradinaru, V. (2024) · Molecular Therapy · doi:10.1016/j.ymthe.2024.04.003 · Cited by 20
“54 The scalable, two-stage HT-APPRAISE strategy we designed provides a path for in silico screening of protein candidates for target receptor binding (Figure 4).”
Hermosilla, A. M., Berner, C., Ovchinnikov, S., & Vorobieva, A. A. (2024) · Protein Science · doi:10.1002/pro.5033 · Cited by 17
“As previously observed (Dauparas et al., 2022; Lin et al., 2023; Wang et al., 2022; Wicky et al., 2022), Alpha-Fold2 could predict the structure of water-soluble designs from a single sequence.”
BackgroundSelf-citationCutting-edge computational approaches in enzyme design and activity enhancement
Sun, R., Wu, D., Chen, P., & Zheng, P. (2024) · Biochemical Engineering Journal · doi:10.1016/j.bej.2024.109510 · Cited by 20
A Deep SE(3)-Equivariant Model for Learning Inverse Protein Folding
McPartlon, M., Lai, B., & Xu, J. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.04.15.488492 · Cited by 17
ESM ‐scan—A tool to guide amino acid substitutions
Totaro, M. G., Vide, U., Zausinger, R., Winkler, A., & Oberdorfer, G. (2024) · Protein Science · doi:10.1002/pro.5221 · Cited by 19
“Deep learning tools trained on structural information are likely superior predictors in tasks where spatial features are pre-eminent, as in the case of protein-protein interactions (Wang et al. 2022, Heinzinger et al. 2023, Shu et al. 2023, Su et al. 2023).”
Graphormer supervised de novo protein design method and function validation
Mu, J., Li, Z., Zhang, B., Zhang, Q., Iqbal, J., Wadood, A., Wei, T., Feng, Y., & Chen, H. (2024) · Briefings in Bioinformatics · doi:10.1093/bib/bbae135 · Cited by 19
“High recovery has not shown strong correlations with the success rate of wet experiments, since a single residue substitution, which may not cause notable changes in metrics, can nevertheless disrupt the overall structure [31].”
BackgroundAsiteDesign: a Semirational Algorithm for an Automated Enzyme Design
Rodà, S., Terholsen, H., Meyer, J. R. H., Cañellas-Solé, A., Guallar, V., Bornscheuer, U. T., & Kazemi, M. (2023) · The Journal of Physical Chemistry B · doi:10.1021/acs.jpcb.2c07091 · Cited by 18
“25,26 Additionally, de novo protein design techniques are becoming increasingly accurate, 24,27,28 which makes it possible to create tailor-made protein scaffolds.”
BackgroundSuper High-Throughput Screening of Enzyme Variants by Spectral Graph Convolutional Neural Networks
Ramírez-Palacios, C., & Marrink, . J. (2023) · Journal of Chemical Theory and Computation · doi:10.1021/acs.jctc.2c01227 · Cited by 19
Generative design of therapeutics that bind and modulate protein states
Chen, T., Hong, L., Yudistyra, V., Vincoff, S., & Chatterjee, P. (2023) · Current Opinion in Biomedical Engineering · doi:10.1016/j.cobme.2023.100496 · Cited by 20
AI in cellular engineering and reprogramming
Capponi, S., & Wang, S. (2024) · Biophysical Journal · doi:10.1016/j.bpj.2024.04.001 · Cited by 20
Toward deep learning sequence–structure co-generation for protein design
Wang, C., Alamdari, S., Domingo-Enrich, C., Amini, A. P., & Yang, K. (2025) · Current Opinion in Structural Biology · doi:10.1016/j.sbi.2025.103018 · Cited by 22
Deep Generative Design of Epitope-Specific Binding Proteins by Latent Conformation OptimizationPreprint
Eguchi, R. R., Choe, C. A., Parekh, U., Khalek, I. S., Ward, M. D., Vithani, N., Bowman, G. R., Jardine, J. G., & Huang, P. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.12.22.521698 · Cited by 21
“In recent work, denoising diffusion probabilistic models [19] have gained attention in protein design [20, 21, 22], and have been shown as being able to generate static helical binders against hotspots by using a variant of RoseTTAFold fine-tuned as a denoising network [23].”
BackgroundExploring the Protein Sequence Space with Global Generative Models
Romero‐Romero, S., Lindner, S., & Ferruz, N. (2023) · Cold Spring Harbor Perspectives in Biology · doi:10.1101/cshperspect.a041471 · Cited by 21
Fang, M., Wang, C., Shi, J., Lian, F., Jin, Q., Wang, Z., Zhang, Y., Chen, P., Cui, Z., Wang, Y., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.11.08.686881 · Cited by 20
Opportunities for engineering outer membrane vesicles using synthetic biology approaches
Kelwick, R. J. R., Webb, A. J., & Freemont, P. S. (2023) · Extracellular Vesicles and Circulating Nucleic Acids · doi:10.20517/evcna.2023.21 · Cited by 17
Rational design of small‐molecule responsive protein switches
Shui, S., Buckley, S., Scheller, L., & Correia, B. E. (2023) · Protein Science · doi:10.1002/pro.4774 · Cited by 19
“Alternatively, inpainting was trained to fill in both sequence and structure of missing parts of a protein through its network that was trained on masked protein segments (Wang et al., 2022).”
BackgroundSelf-citationClick, Compute, Create: A Review of Web‐based Tools for Enzyme Engineering
Tripp, A., Braun, M., Wieser, F., Oberdorfer, G., & Lechner, H. (2024) · ChemBioChem · doi:10.1002/cbic.202400092 · Cited by 17
Soleymani, F., Paquet, E., Viktor, H. L., & Michalowski, W. (2024) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2024.06.021 · Cited by 19
Ding, Y., Zhang, S., Kong, X., Hess, H., & Zhang, Y. (2025) · Advanced Science · doi:10.1002/advs.202500859 · Cited by 25
De novo protein fold design through sequence-independent fragment assembly simulations
Pearce, R., Huang, X., Omenn, G. S., & Zhang, Y. (2023) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2208275120 · Cited by 18
“Meanwhile, the resulting protein folds were largely randomized depending on the stochastic process of the design iterations, where the method was further extended to allow for the incorporation of specific functional sites or structural motifs (Smotifs) (26).”
MethodProteus: Exploring Protein Structure Generation for Enhanced Designability and EfficiencyPreprint
Wang, C., Qu, Y., Peng, Z., Wang, Y., Zhu, H., Chen, D., & Cao, L. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.02.10.579791 · Cited by 28
“Recent advancements in protein structure prediction methods, such as AlphaFold2 (Jumper et al., 2021) and RosettaFold (Baek et al., 2021), have enabled the ’hallucination’ approach (Anishchenko et al., 2021; Wang et al., 2022) to directly generate protein sequences using backpropagation on the…”
MethodIn Silico Approaches for Prediction of Anti-CRISPR Proteins
Makarova, K. S., Wolf, Y. I., & Koonin, E. V. (2022) · Journal of Molecular Biology · doi:10.1016/j.jmb.2023.168036 · Cited by 16
Metadata
- Authors
- Wang, J., Lisanza, S., Juergens, D., Tischer, D., Watson, J. L., Castro, K. M., Ragotte, R., Saragovi, A., Milles, L. F., Baek, M., Anishchenko, I., Yang, W., Hicks, D. R., Expòsit, M., Schlichthaerle, T., Chun, J.-H., Dauparas, J., Bennett, N., Wicky, B. I. M., Muenks, A., DiMaio, F., Correia, B., Ovchinnikov, S., & Baker, D.
- Affiliations
- University of Washington (US)
- École Polytechnique Fédérale de Lausanne (CH)
- Harvard University (US)
- Howard Hughes Medical Institute (US)
- Corresponding author
- Wang, Jue
- Lisanza, Sidney
- Juergens, David
- Tischer, Doug
- Watson, Joseph L.
- Ovchinnikov, Sergey
- Baker, David
- ORCID
- Wang, Jue
- Lisanza, Sidney
- Juergens, David
- Tischer, Doug
- Watson, Joseph L.
- Castro, Karla M.
- Ragotte, Robert
- Saragovi, Amijai
- Milles, Lukas F.
- Baek, Minkyung
- Anishchenko, Ivan
- Yang, Wei
- Hicks, Derrick R.
- Expòsit, Marc
- Schlichthaerle, Thomas
- Chun, Jung-Ho
- Dauparas, Justas
- Bennett, Nathaniel
- Wicky, Basile I. M.
- Muenks, Andrew
- DiMaio, Frank
- Correia, Bruno
- Ovchinnikov, Sergey
- Baker, David
- Venue
- Science
- Published
- 2022-07-22
- Volume
- 377
- Issue
- 6604
- Pages
- 387-394
- ISSN
- 0036-8075 (ISSN-L)
- 1095-9203
- PMC
- PMC9621694
- OpenAlex
- W4286488966
- Cited by
- 487
- References
- 94
- Citations by year
- 2026: 84 · 2025: 107 · 2024: 124 · 2023: 121 · 2022: 49
- FWCI
- 32.74
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Licence
- cc-by-nc-nd
- Locations
- Science (publishedVersion)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- PubMed Central (acceptedVersion, cc-by)
- Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion, cc-by-nc-nd)
- MPG.PuRe (Max Planck Society) (submittedVersion)
- Seoul National University Open Repository (Seoul National University) (submittedVersion)
- Funders
- Howard Hughes Medical Institute
- Washington Research Foundation
- Amgen
- Open Philanthropy Project
- Fundación Bancaria Caixa d'Estalvis i Pensions de Barcelona
- National Institutes of Health (DP5OD026389, HHSN272201700059C)
- National Institute on Aging (U19AG065156)
- National Cancer Institute (R01CA240339)
- National Institute of Allergy and Infectious Diseases (HHSN272201700059C)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
- Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
- Keywords
- protein scaffolding, protein design, deep learning, inpainting, RoseTTAFold, immunogen, metalloproteins, enzyme design, protein structure prediction
- MeSH terms
- Deep Learning, Binding Sites, Catalysis, Protein Binding, Proteins, Protein Engineering, Protein Structure, Secondary, Protein Folding
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.