Protein Folding: Levinthal's Paradox

A protein cannot find its native structure by blindly testing every possible conformation.

Worked example — Protein Folding: Levinthal's Paradox

Possible conformations	5.15 × 10⁴⁷
Exhaustive search time	1.63 × 10²⁷ years
Times the universe's age	1.18 × 10¹⁷×
Fraction sampled before folding	1.94 × 10⁻³⁸

These are the values the calculator shows when the page opens. Change any input above to recompute them.

Loading interactive simulation...

The paradox rules out the algorithm 🖖

Levinthal's estimate was not a proposal for how proteins fold. It is a reductio: even impossibly fast blind sampling cannot cover conformational space, so physical interactions must guide the chain through a tiny, structured part of it.

Below about twenty residues, blind search is fine 🖖

At three states per residue and 10⁻¹³ s a try, a 21-residue chain covers its entire conformational space in a millisecond — a real folding time. Thirty residues takes 20 seconds; fifty takes about 2,300 years. Every ten residues multiplies the search by 3¹⁰, roughly 59,000. The paradox is a statement about size.

The number of states is not the lever 🖖

Allow each junction only two orientations — a coin flip, and physically absurd — and a 100-residue chain still has 10³⁰ conformations, needing 29% of the age of the universe to sample. Cutting the base from three to two buys a factor of 10¹⁷·⁶, once. Every further ten residues costs 59,000, again and again. The residue count is the exponent.

Problem solved in full

A 100-residue protein, three shapes per residue, and a search that outlasts the universe 6 steps
A chain of 100 residues can take three orientations at each one, and it can try a new arrangement every 10⁻¹³ s. Work out how long an exhaustive search takes. Then find the chain length where blind search stops being possible, and say what the paradox actually rules out.
1. Count the arrangements. Three choices at each of 100 residues is 3¹⁰⁰, and in logs that is 100 × log₁₀ 3 = 100 × 0.4771 = 47.71, so 5.15 × 10⁴⁷ conformations.
  
  $lo g_{10} 3^{100} = 100 lo g_{10} 3 = 100 \times 0.4771 = 47.71$
2. Each try takes 10⁻¹³ s, which is about the fastest anything moves in a molecule. Multiplying, the search takes 10⁴⁷·⁷¹ × 10⁻¹³ = 10³⁴·⁷¹ s.
  
  $t = 1 0^{47.71} \times 1 0^{- 13} s = 1 0^{34.71} s$
3. Convert to years by dividing by 3.156 × 10⁷ s. The exponent drops by 7.50, giving 10²⁷·²¹ years.
  
  $\frac{1 0 ^{34.71}}{3.156 \times 1 0 ^{7}} = 1 0^{27.21} yr$
4. The universe is 1.38 × 10¹⁰ years old, which is 10¹⁰·¹⁴. So an exhaustive search takes 10¹⁷·⁰⁷ times the age of the universe — and a real protein of this size folds in something between a millisecond and a second.
  
  $\frac{1 0 ^{27.21}}{1.38 \times 1 0 ^{10}} = 1 0^{17.07} universes$
5. Now find where the trouble starts. Hold three states per residue and walk the chain length up: at 28 residues the search first passes one second, and at 65 it first passes the age of the universe. Below about 20 residues blind search is genuinely fine, which is why short peptides are not paradoxical.
  
  $n = 28 \Rightarrow t > 1 s, n = 65 \Rightarrow t > t_{universe}$
6. Test the obvious escape. If the problem is three states, try two: 2¹⁰⁰ is 10³⁰·¹⁰, and the search still takes 10⁹·⁶ years. Cutting the states per residue by a third removed 17 orders of magnitude and left 9 more than a lifetime.
  
  $lo g_{10} 2^{100} = 30.10 \Rightarrow 1 0^{9.6} yr$
Answer

10²⁷ years, which is 10¹⁷ universes, for something that happens in under a second. The paradox does not say folding is impossible; it says the ALGORITHM is impossible, and those are different claims. Blind search is ruled out, and no amount of tuning the parameters rescues it, because the count is exponential in the chain length and every lever you have is inside the base. What is left is that the protein never enumerates: the energy landscape is a funnel, so almost every downhill step is progress and the chain is steered rather than searching. The sketch above is that funnel — the wide top is the 5.15 × 10⁴⁷, the point at the bottom is the native state N, and the two descending paths arrive together because the route does not matter. Levinthal published this in 1969 as an objection, and it became the argument for the mechanism that replaced it.

Learning path

Computing with molecules

Leads to Strand displacement gate why a physical substrate is not an exemption from complexity.

References (1)

A quantitative treatment of the paradox and why energetic bias collapses the search time: R. Zwanzig, A. Szabo and B. Bagchi, “Levinthal’s paradox.” Proceedings of the National Academy of Sciences 89(1), 20–22, 1992.

By the 42Math team · Updated 2026-08-22

Reviewed against published sources. Corrections welcome. Editorial policy · Corrections

All Bio, Geo, Chem lessons →

Example problems

classic 100 residues - The classic estimate: 5.15 × 10⁴⁷ conformations and 1.63 × 10²⁷ years to try them all.
small fast folder - Thirty residues with two states each is only 1.07 × 10⁹ conformations — small enough that blind search would actually finish.
two torsions per residue - Nine states per residue exactly doubles the exponent: 2.66 × 10⁹⁵ conformations.