About

Watch a protein being made, and fold

Fold Proteins is an interactive model of the moment a protein comes into being: amino acids joined one at a time inside a ribosome, the growing chain sliding out of the exit tunnel, and a physics engine letting it find a shape.

What it is for

Textbooks show protein synthesis as a diagram and protein folding as a before-and-after picture. This app puts the two together in one moving scene, so that the ideas behind them can be seen rather than memorised: that a protein is a chain of 20 kinds of building block; that it grows head first, through a tunnel, at a few amino acids per second; that it starts folding while it is still being made; and that the fold is driven by a handful of simple, weak forces, above all the urge of oily side chains to hide from water.

Goals

• Make the chemistry visible. Every atom of every amino acid is drawn; you can see the water molecule leave when a peptide bond forms, the flat peptide plane, the ring of proline, the charge on lysine.
• Make time visible. Build at 0.1–2 amino acids per second and watch what happens between additions: the chain is never still.
• Let you experiment. Switch forces on and off, change the temperature, compare a fast and an accurate physics library, build one domain of a protein on its own, colour by type or domain.
• Use real proteins. A library of 20 human proteins with sequences from UniProt and domains from the EBI, plus any protein you can find in the AlphaFold database.
• Explain as you go. The Learn pages give the background in plain language, with sources, and link back into the app.

How to use it

1. Pick a protein in the toolbar (or Choose → search AlphaFold), or stay on Custom Amino Acids and add residues by hand from the left panel.
2. Press ▶ Build. A ribosome appears; each new amino acid floats in, docks at the peptidyl transferase centre and bonds to the chain's tail, which slides down the exit tunnel. +1 adds a single residue. The Translation rate slider sets the speed.
3. Watch it fold. When the chain is complete the ribosome fades and the chain is free. Drag to rotate, scroll to zoom, right-drag to pan; the camera follows the chain unless you tell it not to.
4. Experiment. The Physics tab chooses the library and its forces; Appearance sets colours, labels and the ribosome's look; Protein shows the sequence, secondary-structure prediction and domains. Load jumps straight to a pre-folded chain where one exists.

Fold Proteins (foldproteins.org) is an educational tool and not a research instrument; see Scope & limitations. The 3D engine it is built on, FattyInteractModel3D, was first written for a fatty-acid model and is included here unchanged apart from the peptide terms described under Folding Physics.

Scope & Limitations

Built to teach, not to predict

Everything in the app is simplified on purpose. The simplifications are chosen so that the ideas come across (what a peptide bond is, why chains fold, what the ribosome does) at the cost of numerical accuracy. Here is what is real and what is not.

What is real

• The amino acids: every atom, with standard bond lengths and angles ^[1], and the charge each side chain carries at physiological pH.
• The peptide bond: a water molecule leaves, the C–N bond is planar and trans, and the chain grows from the N-terminus to the C-terminus.
• The protein sequences (UniProt) and domain boundaries (EBI Proteins API), and the AlphaFold structures used for the Load and Search AlphaFold features ^{[2, 3]}.
• The order of events in the ribosome: docking at the PTC, the tail sliding down an exit tunnel of about the right length and width ^[4], slower progress for positively charged stretches, folding that begins at the tunnel exit.
• The forces, in kind: hydrophobic attraction, hydrogen bonding, electrostatics, steric repulsion, and a preference for allowed backbone angles ^[5].

What is simplified

• The physics is not molecular dynamics. Neither library solves a real force field. Fast Physics uses geometric constraints plus a few tuned attractions and a Ramachandran prior; Accuracy Physics is the site's general-purpose FattyInteractModel3D engine with the same simplifications. Energies, temperatures and times are in arbitrary units. The shapes that come out are plausible, not predicted.
• No water, no crowding. Water is implied through the hydrophobic term only; the real cell is a dense soup of proteins, ions and membranes.
• No chaperones. Hsp70, chaperonins and the other helpers that shape folding in a living cell are not modelled (yet); every chain folds alone.
• One chain at a time. Multi-chain assemblies, disulfide bridges between chains, aggregation and amyloid cannot form.
• The ribosome is a shape, not a molecule. A smooth tube with a constriction, a PTC marker and two decorative loops, with a uniform wall charge; the real tunnel is lined with specific RNA and protein surfaces, and there is no tRNA, mRNA or elongation factor in the scene.
• Time is slowed and compressed. Real translation runs at 5–6 amino acids per second and a helix turn forms in well under a microsecond; the app builds at 0.1–2 per second and its physics runs at a fixed rate between additions, so the folding you see between two residues is neither to scale nor complete.
• Pre-folded chains are not experimental structures. A prefold is what the app's own physics produced for that protein, so that Load is instant; it is not the PDB or AlphaFold structure, which can be opened side by side in the viewer for comparison ^{[3, 6]}.
• Secondary-structure prediction in the Protein tab is a simple Chou–Fasman-style estimate from sequence, far less accurate than modern predictors.

The numbers in the app's panels (bond lengths, charges, counts of residues) can be trusted; the folds should be read as cartoons of what folding is like. For a research-grade picture of any protein, use its experimental structure in the PDB or its AlphaFold model, both linked from the viewer.

Sources

1. Engh RA, Huber R (1991). Accurate bond and angle parameters for X-ray protein structure refinement. Acta Crystallogr A 47:392-400 · DOI
2. UniProt Consortium (2023). UniProt: the Universal Protein Knowledgebase in 2023. Nucleic Acids Res 51:D523-D531 · DOI
3. Varadi M, Anyango S, Deshpande M, et al. (2022). AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res 50:D439-D444 · DOI
4. Voss NR, Gerstein M, Steitz TA, Moore PB (2006). The geometry of the ribosomal polypeptide exit tunnel. J Mol Biol 360:893-906 · DOI
5. Ramachandran GN, Ramakrishnan C, Sasisekharan V (1963). Stereochemistry of polypeptide chain configurations. J Mol Biol 7:95-99 · DOI
6. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE (2000). The Protein Data Bank. Nucleic Acids Res 28:235-242 · DOI

Folding Physics

How the app folds a chain

The folding you watch is produced by a small physics engine written for this app, not by a research force field. This page explains what it does, how it compares with the real thing, what shapes folding in a cell that it does and does not include, and where its libraries are heading.

The physics in one paragraph

Every atom of the chain is a point with a position and a velocity. Each step (30 per second), forces nudge the atoms, then constraints pull the geometry back to exactly what chemistry allows: every bond length, every bond angle, the flatness of rings and of the peptide bond. This scheme, position-based dynamics, comes from computer graphics (cloth, ropes, soft bodies), where it was invented because it never blows up and runs in real time ^{[1, 2]}. The forces are the few that matter for a fold: a hard steric wall so atoms cannot pass through each other; a hydrophobic attraction between non-polar side chains, standing in for the water that is not simulated; screened electrostatics between charged side chains; backbone hydrogen bonds (N–H toward a C=O at least three residues away) that make helices and sheets; and a φ/ψ prior that keeps each residue's backbone angles in an allowed region of the Ramachandran map. A little random motion ("temperature") keeps the chain exploring, and damping stops it ringing.

How that compares with real simulations

Research-grade molecular dynamics integrates Newton's equations for every atom, including thousands of explicit water molecules, with a force field (AMBER, CHARMM and their relatives) whose hundreds of parameters are fitted to quantum chemistry and experiment ^[3]. The time step is a femtosecond (10⁻¹⁵ s), so a millisecond of folding costs 10¹² steps; it took a purpose-built supercomputer, Anton, to watch a dozen small proteins fold and unfold repeatedly in atomic detail ^{[4, 5]}. Those simulations confirmed what the app assumes: small proteins fold through a few native-like intermediates, and the hydrophobic collapse and secondary structure form together rather than one after the other.

The app is nowhere near that, by design. It has no water, no electronic polarisation, no real energy units and no real clock; its "temperature" is a knob. What it keeps is the logic of folding: real geometry, excluded volume, a hydrophobic drive and hydrogen bonds. That is enough for a chain to collapse, bury its oily residues and form helices where the sequence favours them, in seconds on a laptop, which is what an educational model needs. It is not enough to predict a fold; the structure a chain settles into here is one plausible compact shape among many, not the native one.

One at a time, or all at once

There are two ways to fold a chain in the app, and they give different results on purpose.

Neither route is a search through every shape. Levinthal pointed out in 1969 that a 100-residue chain trying its conformations one by one would need longer than the age of the universe ^[10]; real chains, and the app's, fold down an energy landscape where local structure forms fast and guides the rest ^[11]. Try the same protein both ways and compare: Build the insulin B chain at 0.5 aa/s, then Clear and Load it.

What shapes folding in a cell

The libraries

The physics is not one program but a set of interchangeable libraries (Physics tab → library dropdown), each a small JavaScript module with its own settings and help page:

• Accuracy Physics (default) — the site's general all-atom engine, FattyInteractModel3D, with its full set of terms (Lennard-Jones, Coulomb, bond springs, torsions, hydrophobic band) plus the peptide terms this app adds: protein backbone torsions, polar hydrogens without van der Waals, backbone H-bonds, and the same φ/ψ prior as the fast library. The most chemistry per atom, and the slowest.
• Fast Physics — Amino Folding — the position-based engine described above, built for long chains; it runs in a background thread so the page stays responsive. Builds through the ribosome properly; the one to use for anything over ~100 residues.
• Chou–Fasman — a secondary-structure guide: a per-residue helix / sheet / turn prediction from the 1974 propensity tables ^[17], applied as restraints on top of one of the engines above, for seeing where a sequence "wants" helices.
• No forces, Stretch display and Stretch display — parallel bonds — not folding at all: geometry-only motion, and two lay-it-flat views for reading a chain residue by residue.

Shared by all of them: the Movement settings (temperature, damping, solver passes, step limit) and the Build module, the ribosome tunnel with its wall rules, which attaches to whichever library is building.

Where this is going: open source, improved over time

The goal is for these libraries to become an open-source project: a documented interface (createEngine, buildSettings, a help.md, a settings.json; see physics/README.md) that anyone can write a library against, so that better physics can be dropped in without touching the app. The obvious next steps, roughly in order:

1. A stiffer, cooler Accuracy Physics in the tunnel so that the all-atom library builds as cleanly as the fast one (today its chain coils at the tunnel entrance).
2. Chaperones at the exit: trigger factor / NAC as a damping of hydrophobic collapse for the first residues outside, then an Hsp70-style hold on exposed sticky stretches.
3. Disulfide bonds and metal ions, so insulin, lysozyme and calmodulin can reach their real folds.
4. A better secondary-structure predictor than Chou–Fasman, and a native-contact guide that can pull a chain toward its AlphaFold or PDB structure when you ask it to.
5. Validation: an automatic comparison of what each library produces with the experimental structure (Cα RMSD, contact overlap), published with each release, so improvements can be measured rather than eyeballed.

Contributions, corrections and better parameters are welcome; each library's help page (the ? next to its name in the Physics tab) states exactly what it does.

Good places to learn more about proteins

Sources

1. Müller M, Heidelberger B, Hennix M, Ratcliff J (2007). Position based dynamics. J Vis Commun Image Represent 18:109-118 · DOI
2. Macklin M, Müller M, Chentanez N (2016). XPBD: position-based simulation of compliant constrained dynamics. Proc. 9th Int. Conf. on Motion in Games (ACM) 49-54 · DOI
3. Karplus M, McCammon JA (2002). Molecular dynamics simulations of biomolecules. Nat Struct Biol 9:646-652 · DOI
4. Shaw DE, Maragakis P, Lindorff-Larsen K, et al. (2010). Atomic-level characterization of the structural dynamics of proteins. Science 330:341-346 · DOI
5. Lindorff-Larsen K, Piana S, Dror RO, Shaw DE (2011). How fast-folding proteins fold. Science 334:517-520 · DOI
6. Kramer G, Boehringer D, Ban N, Bukau B (2009). The ribosome as a platform for co-translational processing, folding and targeting of newly synthesized proteins. Nat Struct Mol Biol 16:589-597 · DOI
7. Fedyukina DV, Cavagnero S (2011). Protein folding at the exit tunnel. Annu Rev Biophys 40:337-359 · DOI
8. Cassaignau AME, Cabrita LD, Christodoulou J (2020). How does the ribosome fold the proteome?. Annu Rev Biochem 89:389-415 · DOI
9. Anfinsen CB (1973). Principles that govern the folding of protein chains. Science 181:223-230 · DOI
10. Levinthal C (1969). How to fold graciously. Mössbauer Spectroscopy in Biological Systems (Univ. of Illinois) 22-24 · Link
11. Dill KA, MacCallum JL (2012). The protein-folding problem, 50 years on. Science 338:1042-1046 · DOI
12. Voss NR, Gerstein M, Steitz TA, Moore PB (2006). The geometry of the ribosomal polypeptide exit tunnel. J Mol Biol 360:893-906 · DOI
13. Hartl FU, Bracher A, Hayer-Hartl M (2011). Molecular chaperones in protein folding and proteostasis. Nature 475:324-332 · DOI
14. Balchin D, Hayer-Hartl M, Hartl FU (2016). In vivo aspects of protein folding and quality control. Science 353:aac4354 · DOI
15. Chandler D (2005). Interfaces and the driving force of hydrophobic assembly. Nature 437:640-647 · DOI
16. Levy ED, De S, Teichmann SA (2012). Cellular crowding imposes global constraints on the chemistry and evolution of proteomes. Proc Natl Acad Sci USA 109:20461-20466 · DOI
17. Chou PY, Fasman GD (1974). Prediction of protein conformation. Biochemistry 13:222-245 · DOI
18. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE (2000). The Protein Data Bank. Nucleic Acids Res 28:235-242 · DOI
19. Varadi M, Anyango S, Deshpande M, et al. (2022). AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res 50:D439-D444 · DOI
20. UniProt Consortium (2023). UniProt: the Universal Protein Knowledgebase in 2023. Nucleic Acids Res 51:D523-D531 · DOI
21. Milo R, Phillips R (2015). Cell Biology by the Numbers. Garland Science, New York · Link

References

The research behind the app

The Learn pages cite published sources, listed here. The app currently uses 44 references, published between 1951 and 2023; 42 link to a DOI or PubMed record and 2 more to the publisher's page.

All references

1. Anfinsen CB (1973). Principles that govern the folding of protein chains. Science 181:223-230 · DOI
2. Balchin D, Hayer-Hartl M, Hartl FU (2016). In vivo aspects of protein folding and quality control. Science 353:aac4354 · DOI
3. Ban N, Nissen P, Hansen J, Moore PB, Steitz TA (2000). The complete atomic structure of the large ribosomal subunit at 2.4 Å resolution. Science 289:905-920 · DOI
4. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE (2000). The Protein Data Bank. Nucleic Acids Res 28:235-242 · DOI
5. Boyer PD (1997). The ATP synthase — a splendid molecular machine. Annu Rev Biochem 66:717-749 · DOI
6. Cassaignau AME, Cabrita LD, Christodoulou J (2020). How does the ribosome fold the proteome?. Annu Rev Biochem 89:389-415 · DOI
7. Chandler D (2005). Interfaces and the driving force of hydrophobic assembly. Nature 437:640-647 · DOI
8. Chiti F, Dobson CM (2017). Protein misfolding, amyloid formation, and human disease: a summary of progress over the last decade. Annu Rev Biochem 86:27-68 · DOI
9. Chou PY, Fasman GD (1974). Prediction of protein conformation. Biochemistry 13:222-245 · DOI
10. Ciryam P, Tartaglia GG, Morimoto RI, Dobson CM, Vendruscolo M (2013). Widespread aggregation and neurodegenerative diseases are associated with supersaturated proteins. Cell Rep 5:781-790 · DOI
11. Crick F (1970). Central dogma of molecular biology. Nature 227:561-563 · DOI
12. Dill KA, MacCallum JL (2012). The protein-folding problem, 50 years on. Science 338:1042-1046 · DOI
13. Doyle DA, Morais Cabral J, Pfuetzner RA, Kuo A, Gulbis JM, Cohen SL, Chait BT, MacKinnon R (1998). The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280:69-77 · DOI
14. Engh RA, Huber R (1991). Accurate bond and angle parameters for X-ray protein structure refinement. Acta Crystallogr A 47:392-400 · DOI
15. Fedyukina DV, Cavagnero S (2011). Protein folding at the exit tunnel. Annu Rev Biophys 40:337-359 · DOI
16. Hartl FU, Bracher A, Hayer-Hartl M (2011). Molecular chaperones in protein folding and proteostasis. Nature 475:324-332 · DOI
17. Hershko A, Ciechanover A (1998). The ubiquitin system. Annu Rev Biochem 67:425-479 · DOI
18. Ingolia NT, Lareau LF, Weissman JS (2011). Ribosome profiling of mouse embryonic stem cells reveals the complexity and dynamics of mammalian proteomes. Cell 147:789-802 · DOI
19. Jumper J, Evans R, Pritzel A, et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature 596:583-589 · DOI
20. Karplus M, McCammon JA (2002). Molecular dynamics simulations of biomolecules. Nat Struct Biol 9:646-652 · DOI
21. Kauzmann W (1959). Some factors in the interpretation of protein denaturation. Adv Protein Chem 14:1-63 · DOI
22. Knowles TPJ, Vendruscolo M, Dobson CM (2014). The amyloid state and its association with protein misfolding diseases. Nat Rev Mol Cell Biol 15:384-396 · DOI
23. Kramer G, Boehringer D, Ban N, Bukau B (2009). The ribosome as a platform for co-translational processing, folding and targeting of newly synthesized proteins. Nat Struct Mol Biol 16:589-597 · DOI
24. Kühlbrandt W (2019). Structure and mechanisms of F-type ATP synthases. Annu Rev Biochem 88:515-549 · DOI
25. Levinthal C (1969). How to fold graciously. Mössbauer Spectroscopy in Biological Systems (Univ. of Illinois) 22-24 · Link
26. Levy ED, De S, Teichmann SA (2012). Cellular crowding imposes global constraints on the chemistry and evolution of proteomes. Proc Natl Acad Sci USA 109:20461-20466 · DOI
27. Lindorff-Larsen K, Piana S, Dror RO, Shaw DE (2011). How fast-folding proteins fold. Science 334:517-520 · DOI
28. Macklin M, Müller M, Chentanez N (2016). XPBD: position-based simulation of compliant constrained dynamics. Proc. 9th Int. Conf. on Motion in Games (ACM) 49-54 · DOI
29. Milo R, Phillips R (2015). Cell Biology by the Numbers. Garland Science, New York · Link
30. Müller M, Heidelberger B, Hennix M, Ratcliff J (2007). Position based dynamics. J Vis Commun Image Represent 18:109-118 · DOI
31. Nirenberg MW, Matthaei JH (1961). The dependence of cell-free protein synthesis in E. coli upon naturally occurring or synthetic polyribonucleotides. Proc Natl Acad Sci USA 47:1588-1602 · DOI
32. Nissen P, Hansen J, Ban N, Moore PB, Steitz TA (2000). The structural basis of ribosome activity in peptide bond synthesis. Science 289:920-930 · DOI
33. Noji H, Yasuda R, Yoshida M, Kinosita K (1997). Direct observation of the rotation of F1-ATPase. Nature 386:299-302 · DOI
34. Pace CN, Shirley BA, McNutt M, Gajiwala K (1996). Forces contributing to the conformational stability of proteins. FASEB J 10:75-83 · DOI
35. Pauling L, Corey RB, Branson HR (1951). The structure of proteins: two hydrogen-bonded helical configurations of the polypeptide chain. Proc Natl Acad Sci USA 37:205-211 · DOI
36. Perutz MF, Rossmann MG, Cullis AF, Muirhead H, Will G, North ACT (1960). Structure of haemoglobin: a three-dimensional Fourier synthesis at 5.5-Å resolution, obtained by X-ray analysis. Nature 185:416-422 · DOI
37. Ramachandran GN, Ramakrishnan C, Sasisekharan V (1963). Stereochemistry of polypeptide chain configurations. J Mol Biol 7:95-99 · DOI
38. Rosenbaum DM, Rasmussen SGF, Kobilka BK (2009). The structure and function of G-protein-coupled receptors. Nature 459:356-363 · DOI
39. Shaw DE, Maragakis P, Lindorff-Larsen K, et al. (2010). Atomic-level characterization of the structural dynamics of proteins. Science 330:341-346 · DOI
40. UniProt Consortium (2023). UniProt: the Universal Protein Knowledgebase in 2023. Nucleic Acids Res 51:D523-D531 · DOI
41. Varadi M, Anyango S, Deshpande M, et al. (2022). AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res 50:D439-D444 · DOI
42. Voss NR, Gerstein M, Steitz TA, Moore PB (2006). The geometry of the ribosomal polypeptide exit tunnel. J Mol Biol 360:893-906 · DOI
43. Walker JE (2013). The ATP synthase: the understood, the uncertain and the unknown. Biochem Soc Trans 41:1-16 · DOI
44. Wu G (2009). Amino acids: metabolism, functions, and nutrition. Amino Acids 37:1-17 · DOI

Data sources used by the app

• UniProt — the protein sequences and names in the library and in search results.
• EBI Proteins API — domain, repeat and motif annotations for the Domains panel.
• AlphaFold Protein Structure Database and PDBe — predicted and experimental structures shown in the viewer.
• Engh & Huber (1991) — the bond lengths and angles the amino acid templates are built from.

Source file: data/learn_refs.json. The counts on this page are read from it, so they stay up to date.

Privacy

Nothing about you leaves your browser

Fold Proteins has no accounts, no sign-in, no tracking and no analytics. Everything, including the physics, runs in your browser.

What is stored on your device

The app remembers a few preferences in your browser's local storage so that they are the same next time. They never leave your device and you can clear them at any time by clearing the site's data in your browser.

What is sent to other servers

The app only contacts other servers when you ask it to look something up, and it sends only what is needed for that request:

• Search AlphaFold sends your search text to the EMBL-EBI search service (www.ebi.ac.uk) and, when you pick a result, fetches that protein's sequence from UniProt and its structure from the AlphaFold database (alphafold.ebi.ac.uk).
• Domains for a searched protein are fetched from the EBI Proteins API; for the built-in library they are already on this site.
• The structure viewer (AlphaFold / PDB view) loads the PDBe Mol* viewer from the jsDelivr CDN and structure files from PDBe and AlphaFold.

These are public scientific services run by EMBL-EBI (and a public CDN); their own privacy notices apply to the requests they receive, which carry your IP address like any web request. No search text or sequence is sent anywhere else, and nothing is sent at all when you use the built-in library offline from these services.

Cookies

The site sets no cookies. Session storage is used only to remember whether you closed the notice under the top bar and, when you arrived through a link with ?ref_from=, where to go back to.

The site is hosted on Cloudflare Pages, which, like any host, logs requests (IP address, user agent) for operation and abuse prevention under its own privacy policy. Fold Proteins itself keeps no logs.