Claim № 4.03  · Pillar 4 · Case file

“Functional protein folds are common enough for random mutation to find them.”

Disciplines: Biochemistry · Information theory
Status: draft · pending expert review
UnderExaminationTwo claims · two grades

§1 The claim, as taught

New proteins arise as mutations alter amino acid sequences; natural selection preserves those that improve function. Over time this process explores the range of possible proteins and has produced the diversity of protein structures seen today.Representative summary of how protein evolution is presented in introductory texts (e.g. Campbell Biology; Alberts Molecular Biology of the Cell). PLACEHOLDER: to be replaced with exact quotations and page numbers.

§2 The strongest evidence offered

These are the main lines of evidence that working proteins are within reach of random search. We present each one fairly and say what it actually shows.

Observed

Function in random sequences

From a library of random 80-residue sequences, Keefe and Szostak selected proteins that bind ATP. They estimated that roughly 1 in 1011 random sequences has comparable binding activity.
Keefe & Szostak, Nature 410 (2001)
Inferred

Genes from non-coding DNA

Yeast translates many short open reading frames outside known genes. Carvunis and colleagues proposed that some of these “proto-genes” are on their way to becoming new genes.
Carvunis et al., Nature 487 (2012)
Observed

Proteins tolerate many substitutions

Mutating λ repressor showed that many surface positions accept a wide range of amino acids, while buried positions accept few. The authors read this as substantial flexibility in the sequences that specify a fold.
Reidhaar-Olson & Sauer, Proteins 7 (1990)
Inferred

Latent, promiscuous functions

Many enzymes have weak side-activities, and their structures shift between states. Tokuriki and Tawfik argued this “dynamism” makes proteins readily evolvable toward new functions.
Tokuriki & Tawfik, Science 324 (2009)

§3 The grade: two claims hiding in one

Two different claims are often run together here: that *simple* functions such as binding can appear in random sequences, and that *complex enzymatic folds* can. Graded separately: Hover or tap a rung to see why it got its score. How the grading works →

Simple functions such as binding occur at measurable rates in random sequences

BStrongly supported
Observation
///
Directly selected and measured from random libraries.
Hypothesis
///
Quantitative: a stated frequency for a defined activity.
Prediction
//
Consistent with expectations from earlier selection experiments.
Experiment
///
Controlled in vitro selection.
Replication
//
Similar approaches have found other binders; methods vary between labs.
Falsifiability
//
Alternative explanations (contamination, artefacts) were addressed.
Method score 15/18

Complex enzymatic folds are common enough to be found by undirected search

DExtrapolation
Observation
/
Estimates for enzymatic folds rest on extrapolation from small samples; Axe's estimate for a β-lactamase domain is about 1 in 10^77^.
Hypothesis
//
Quantitative and therefore testable in principle.
Prediction
/
No prediction of a specific new enzyme fold has been confirmed.
Experiment
/
Binding studies don't test catalysis, and catalysis studies give far lower prevalences.
Replication
/
Few independent measurements for enzymatic folds, using different methods.
Falsifiability
//
A clear measurement could settle it either way. That is to the claim's credit.
Method score 8/18
ADemonstratedBStrongly supportedCPlausible inferenceDExtrapolationEAssumption

§4 What it would take

Here is what would move the claim up a grade. These are not impossible demands. They are the observations the claim itself implies.

  • Measure enzymes, not only bindersMeasure the prevalence of a working enzymatic fold, by more than one lab, with methods that can be compared.
  • Show it works in a cellShow that sequences selected from random libraries fold stably and do their job inside a living cell.
  • Close the arithmeticShow that the number of sequences life could have tried in Earth's history exceeds the search the measured rarity requires.

§5 Another reading of the same data

The headline numbers are not measuring the same thing. Keefe and Szostak measured how often random sequences bind a small molecule. Axe estimated how often sequences fold and catalyse like a specific working enzyme. Binding is the easier task by far, and the selected binders were then improved through further rounds of mutation and selection.

Axe's figure of roughly 1 in 1077 is disputed. Critics argue that one enzyme's sensitivity can't be generalised to all folds, and that evolution need not find any particular fold, only some working one. Those are fair points, and this card is graded as a draft for exactly that reason. But even generous estimates leave a search problem: the number of possible sequences for a modest 150-residue protein is 20150, about 10195.

Durston and colleagues proposed a way to measure the functional information in protein families directly from their sequence variation. On their measures, many protein families carry hundreds of bits of functional information. Design proponents argue that no undirected process has been observed to generate that much; that is their inference, not Durston's measurement, and it is the kind of claim the Information lens will examine.

Read through a design lens, protein folds look like a set of well-chosen solutions in a vast space of non-solutions, found not by search but by knowledge. That is a claim about causes, and it should be tested like any other. The questions below are a place to start.

§6 Reasoning check

The grade weighs the evidence. This check looks at the reasoning: the step from the evidence to the conclusion. The same tests are applied to both readings. How the reasoning check works →

In the claim as taught
  • Clear case

    From proteins that tolerate substitutions, and binders improved by selection, to the origin of every protein fold. The claim as taught presents the conclusion as a description of what happened.

  • Clear case
    In the evidence: Function in random sequences

    Binding a small molecule is counted as “function,” and then set beside estimates for folded enzymes. The two numbers measure different things.

In the other reading
  • Possible

    Axe's figure of about 1 in 1077 was measured for one enzyme domain. Applied to all folds, it generalises from a single case. The card says this openly; the argument still leans on the number.

  • Possible

    “No undirected process has been observed to generate that much functional information” is an absence of evidence. It counts only if the search has been thorough enough that such a process should have been seen.

§7 Questions to ponder

  • Binding ATP and catalysing a reaction are different tasks. Which one does the theory need?
  • Why might estimates of functional sequences differ by more than sixty orders of magnitude?
  • If proto-genes are short and weakly functional, what would it take to show one becoming a complex enzyme?

§8 Sources

Supporting the mainstream claim
  1. Keefe AD, Szostak JW. “Functional proteins from a random-sequence library.” Nature 410:715–718 (2001).C
  2. Carvunis A-R, et al. “Proto-genes and de novo gene birth.” Nature 487:370–374 (2012).C
  3. Reidhaar-Olson JF, Sauer RT. “Functionally acceptable substitutions in two α-helical regions of λ repressor.” Proteins 7:306–316 (1990).B
  4. Tokuriki N, Tawfik DS. “Protein dynamism and evolvability.” Science 324:203–207 (2009).C
Critical & alternative perspectives
  1. Axe DD. “Estimating the prevalence of protein sequences adopting functional enzyme folds.” J. Mol. Biol. 341:1295–1315 (2004).C
  2. Durston KK, Chiu DKY, Abel DL, Trevors JT. “Measuring the functional sequence complexity of proteins.” Theoretical Biology and Medical Modelling 4:47 (2007).C
  3. Meyer SC. Signature in the Cell. HarperOne (2009).C

Source grades are provisional. In production each links to a full published evaluation that readers can challenge.

§9 Your notes

The literature for this pillar →