Biology EP. 05 / 64 By SENSMAN

Levinthal's Paradox — How Proteins Fold So Fast

A protein is not a new substance being manufactured. A single string folds, and the moment it has a shape it has a property. No blueprint, no supervisor.

How a protein gets its shape — an amino acid chain folding into a globular protein

Last time we counted the lodgers sharing this body.

Now back inside the cell.

A billion and a half quail eggs

Take a cell 10 micrometres across and make it a dome 50 metres wide again. Five million times.

The dome is packed with things the size of quail eggs.

How many? “Hundreds of millions” is the usual answer. It is short by an order of magnitude.

Protein density runs at 2–4 million molecules per cubic micrometre of cell volume, and it holds roughly steady whether the cell is E. coli, yeast or human (Milo, 2013).

A sphere 10 micrometres across has a volume of about 524 cubic micrometres. Multiply:

about 1 to 2 billion

A billion and a half quail eggs in one dome.

Not dust — a grain of millet

These eggs are not solid lumps. Peel one and there is a single thread inside, folded and coiled several times over, holding its own position.

The smallest piece of that thread is an amino acid.

At the same scale throughout:

WhatReal size× 5,000,000
Cell10 μm acrossa dome 50 m wide
Folded globular protein4–5 nm across2–2.5 cm · a quail egg
One amino acid0.6–0.8 nm3–4 mm · a grain of millet
Peptide (10 residues, extended)~3.5 nm1.8 cm
Polypeptide (300 residues, extended)~105 nm52 cm

A single amino acid is larger than dust. It is a grain of millet.

It has form and no direction. Alone it does nothing.

52 centimetres folds into 2

A fully extended chain gains about 0.35 nanometres per residue. Three hundred amino acids make 105 nanometres — 52 centimetres at this scale.

Fold it and it becomes a ball 4 nanometres across. Two centimetres.

52 cm of string → a 2 cm ball. The length drops to a twenty-sixth.

A protein is not a new substance being manufactured. It is a substance whose properties are set the moment it has a shape.

Same material, same length. Fold it one way and it is an enzyme, another and it is scaffolding, another and it carries a signal.

What tells a protein how to fold?

Is there a drawing somewhere that specifies the fold?

Anfinsen denatured ribonuclease and then washed the denaturant away. The chain returned to its original structure on its own and recovered its activity. Nothing was added from outside.

The information needed to fold sits in the sequence itself. The 1972 Nobel Prize in Chemistry went to that conclusion.

Each amino acid has an affinity for water, a direction of charge, a volume. Those properties settle wherever the energy is lowest.

No blueprint. No supervisor.

What is Levinthal’s paradox?

In 1969 Levinthal did the counting.

A chain of 100 amino acids. Allow only three arrangements per residue:

Possibilities3^100 ≈ 10^47
Time to sample one10^-13 s
Time to sample allabout 10^27 years
Age of the universe13.8 billion = 1.4 × 10^10 years

A hundred quadrillion times the age of the universe.

Real folding finishes between microseconds and seconds.

So a protein does not find its answer by going through the options. It slides down an energy landscape shaped like a funnel. Wherever it starts, the floor is mostly the same place.

It is not searching. It is falling.

No supervisor, but there are bodyguards

Protein concentration in the cytosol runs at 200–300 grams per litre. At that density a chain fresh off the ribosome is likely to clump with its neighbour before it can fold.

So chaperones attach. Hsp70 grips exposed hydrophobic patches and keeps them from tangling; GroEL/GroES builds a barrel-shaped chamber, shuts one chain inside and lets it fold alone.

A chaperone does not tell the chain how to fold. It clears a space where nothing interferes.

Anfinsen’s conclusion holds. It is only that inside a crowded cell, being left alone takes machinery.

Is the cell simply protein?

Metabolism does not start without enzymes. Cell shape is not held without structural proteins. External signals are not registered without receptors.

All true. Push it to the end, though, and something catches.

In the ribosome — the machine that makes protein — there is no protein within about 18 ångströms of the site where the peptide bond is actually formed. RNA makes the bond (Nissen et al., 2000).

The heart of the machine that stamps out every protein is not protein.

So the accurate version reads: most of what a cell does resolves into changes of protein state, and the catalyst that makes those proteins is RNA.

The ones that do not fold

The rule that shape sets property has exceptions.

About a third of eukaryotic proteins carry disordered stretches longer than 30 residues (Ward et al., 2004). Those stretches do not settle into one structure. They take whatever shape the partner calls for.

Having no fixed shape is the function, not the defect. Regulatory proteins that must bind many partners tend to have the longest disordered regions.

Same sequence, different fold

And there is a worse exception.

The prion protein is helix-dominated in its normal state. The same sequence folded sheet-dominated causes disease.

The sequence is unchanged. Only the fold differs.

Then it gets worse. The misfolded form grabs the normal form and converts it to the same shape. It transmits with no nucleic acid at all, by shape alone (Prusiner, 1982).

Shape is information. In this one case that is not a metaphor.

Still unsolved

In 2020 AlphaFold2 predicted structures from sequence alone, at an accuracy hard to separate from experiment.

Prediction is settled for practical purposes. The folding route is not.

In what order the chain collapses, and why that route is so fast, remains open. We have the answer sheet and are still looking for the working.

In short

  • A cell 10 μm across holds 1–2 billion protein molecules. “Hundreds of millions” is an order of magnitude short
  • At 5 million times, an amino acid is a 3–4 mm grain; a folded protein is a 2 cm quail egg
  • A 300-residue chain is 52 cm extended and 2 cm folded — a twenty-sixth of the length
  • Folding information is in the sequence. No blueprint, no supervisor
  • Exhaustive search would take 10^27 years. The chain falls; it does not search
  • Chaperones do not teach folding. They isolate the chain so it cannot clump
  • In the ribosome, the peptide bond is made by RNA, not protein
  • A third of eukaryotic proteins work without folding
  • Prions share a sequence and differ only in fold — and that shape is infectious

This piece is rebuilt for search readers from “Protein” of the serial essay The Great Corona Swindle. The original’s point of view is kept; the sizes and counts have been reset against the latest estimates.

Frequently asked

What is the difference between a peptide, a polypeptide and a protein?
They are the same chain at different lengths and states. A few to a few dozen linked amino acids is a peptide; dozens to hundreds or more is a polypeptide. When that polypeptide folds into its own three-dimensional structure and acquires a function, it is a protein. The boundaries are convention rather than strict definition.
What tells a protein to fold?
Nothing does. Anfinsen showed that ribonuclease, denatured and then freed of the denaturant, recovers its original structure and activity on its own. The information needed to fold is in the amino acid sequence itself. Cells do contain chaperone proteins, but they do not dictate the fold — they prevent the chain from clumping.
What is Levinthal's paradox?
A chain of 100 amino acids with only three arrangements per residue has 3^100 ≈ 10^47 possibilities. Sampling one every 10^-13 seconds, an exhaustive search would take 10^27 years. Real folding takes microseconds to seconds. So proteins do not fold by searching the space.
Did AlphaFold solve the protein folding problem?
It solved structure prediction to a practical standard; at CASP14 in 2020 its median accuracy was close to indistinguishable from experimental structures. How the chain reaches that structure, and why the route is so fast, is a separate question and still open.

References

#protein#amino acid#peptide#polypeptide#protein folding#Levinthal's paradox#Anfinsen#chaperone#prion#AlphaFold

#단백질#아미노산#펩타이드#폴리펩타이드#단백질접힘#레빈탈의역설#안핀센#샤페론#프리온#알파폴드

Published September 25, 2026