Detwiler: Questions Behe, Polyphen, and Ratchets

If I’m doing this wrong, please forgive me, but this is the only open thread I could find close to the whole current Behe controversy. I checked the PolyPhen software and since 2012 (the paper in question was published in 2014), besides the results of “benign”, “possibly damaging”, and “probably damaging” results, there is also a possible result of “unknown” under which I assume increased thermal stability or maybe stronger binding affinities might be found. Is this significant? One of the more common responses to the ID claim that it is impossible to eliminate deleterious mutations because of a muller’s ratchet type phenomenon is that selection would favor loci that combine the beneficial mutations. Even though it seems that Behe cherry picked that data here, even the original table shows no “unknown” results.


Great questions @John_Detwiler. Welcome to PS!

What are your thoughts @evograd, @davecarlson, and @T_aquaticus?

1 Like

From my limited understanding, those algorithms can only calculate the probability of a mutation being deleterious based on known disease causing mutations. I don’t think the algorithms are capable of predicting an increase in thermal stability or binding affinities, much less beneficial mutations. And I will repeat the caveat from before, my knowledge of these algorithms is very limited so please don’t take this as gospel.

Muller’s ratchet is much, much stronger in haploid asexual species. Sexual reproduction allows a shuffling of mutations through time which helps them avoid Muller’s ratchet. Therefore, that ID claim would mostly apply to prokaryotes.


From the Polyphen-2 manual (emphasis added):

UNKNOWN prediction
Sometimes PolyPhen-2 reports “This mutation is predicted to be UNKNOWN (score is not available)” as a prediction outcome. What PolyPhen2 is trying to say is that predicting the substitution’s effect was not possible in this particular case due to lack of multiple sequence alignment. The issue was largely addressed in
PolyPhen-2 v2.2.2, thanks to integration of MultiZ genomic multiple alignments. This
allowed for expanding prediction coverage significantly, especially in non-globular domains. Overall, close to 95% of all sequence positions in known UniProtKB proteins can
be now successfully classified. However, you may still encounter UNKNOWN predictions
in rare cases.
The most likely reason for such reports is if much of the sequence of your protein of
> interest is non-alignable due to the presence of large stretches of repeats and/or high compositional biases. For known proteins, you can easily check this by browsing the UniProt
sequence annotations for the protein: UniProt.
Such sequence features often make it impossible to search for homologous sequences,
build reliable multiple alignments, and ultimately infer conservation scores. The issue
affects many non-globular proteins, including collagens, matrix proteins, DNA/RNAbinding proteins, muscle proteins, and more.
You can see this for yourself in the PolyPhen-2 Web interface by clicking on the [+] icon
next to the “Multiple sequence alignment” label to inspect the MSA for your protein.
You can also start an interactive MSA browser via the link at the bottom of the in-line
alignment viewer section (requires Java browser plug-in).

Based on this information, it doesn’t seem that results falling in the “unknown” category would suggest higher binding affinities or other putatively positive phenotypic effects.


it seems everyone discussing this is always using the term beneficial to apply to morphological effect, something which Behe affirms and is in fact relying on in his rule. The other metric seems to be the integrity of the protein with itself and the bonds it makes with other proteins. Obviously this cannot always continue to degrade. I’m looking to make the strongest case for evolutionists, so to me it seems that more sustainable bonds might be found under the unknown category. Benign results might make an umbrella for shifting function, but still not de novo.

1 Like

That isn’t correct @John_Detwiler. We are not using “beneficial” to refer to morphologic effect.

We aren’t evolutionists really, so…

In evolutionary science, we don’t think functions arise without any precursors. So, this isn’t a problem. Do you have an evidence of things that arose without precursors? I’ve looked all over and haven’t found any.

for one, I don’t see a reason that any of the results might not have positive phenotypic effect. I don’t think Behe is focusing on phenotype, he’s saying positive phenotypic effect is what cements the genotypic effects, which he claims are almost always negative from a stability perspective. I’m trying to imagine what sustainable genotypic effects would look like given the allowable results of this tool. Obviously “slightly beneficial” is not even one of the results possible in the program, so I’m trying to see how stronger bonds might be classified. Anything unknown seems to at least allow for that. I read somewhere that a regression analysis was done on the number of new folds discovered over time that gave great results with high confidence. I don’t remember where it is though :frowning:

1 Like

Yes, and the evidence he presents is from PolyPhen which does not demonstrate that there was a negative genotypic effect in any way:

Behe does think that mutations (essentially) always destabilize a protein, but this is not established with evidence. It is merely an assertion, countered by immense amounts of evidence.

There is software (not polyphen) that can estimate the stability of proteins. Would it be important or interesting to you to see what that software tells us about these mutations? Would it make a difference if it showed that none of these mutations were strongly predicted to destabilize the protein?

To be clear, did you see the original table?

I confess that I really didn’t follow what you are saying here. Behe (and polyphen-2) are absolutely focusing on the phenotype. Proteins are phenotypes!

Obviously “slightly beneficial” is not even one of the results possible in the program, so I’m trying to see how stronger bonds might be classified. Anything unknown seems to at least allow for that.

As I mentioned above, the classification of “unknown” simply means that there is no data with which to make any sort of prediction regarding functional effect, most often because the variant in question could not be successfully aligned with the reference sequence. An “unknown” result doesn’t tell us anything.

1 Like

@John_Detwiler, just so this isn’t missed, you asked a good quesiton that was answered correctly by @davecarlson.

You are asking a different question that Behe,

I think you are confusing “beneficial” with “protein stability.” We can study the protein stability of these mutations with other programs.

@davecarlson that isn’t precisely right.

Behe is arguing that there is a disconnect between morphological fitness and biochemical complexity. Damaging genes, in his view, can increase organismal fitness, and is the most common way of increasing fitness. So his argument depends on (among many other things) treating these two as different and objectively measureable things.

As we have seen (and Lenski explains), polyphen doesn’t tell him what he needs it to tell him (other than as a quirk using "damaged). He thinks it tells us if the mutation “damages” a protein or not, but it doesn’t. Yes, it is fairly shocking, that Behe rests the premise of his book on something this incorrect, but he does.

@John_Detwiler, have you read this article yet?

1 Like

To be more precise I should say phenotypic effect.

but you just said aren’t really an evolutionist right? :smiley: So who is “we”?
I’m just looking to build the case for climbing Mt. Improbable. I’m trying to characterize what results might look like given the software that was used.

1 Like

Good point. I wasn’t thinking about the broader context.

1 Like

Doesn’t change anything I wrote, and the data and reasoning equally undermines Behe’s hypothesis.

I’m a Christian that follows Jesus. My worldview does not rest on evolution, but on Jesus. I am not an evolutionist. Many of the atheists here too, do not rest their worldview on evolution. Many of us affirm evolutionary science, but that doesn’t make us “ists.”

Given the software used, the results should look just like it does. There is not one thing in that Polar Bear paper that challenges evolutionary science, and quite a bit that supports it. Behe just misread the paper. It is hard to help someone who just misread the clear plain text of the paper.

I’m not sure of a better word for the outward effects of an SNP visible to natural selection than “phenotype”. Yes, obviously there is a context in which a protein itself is a phenotype of it’s own DNA sequence. Behe’s rule is not about that per se. My understanding is that it is possible for a person to see two different metrics, one that describes the strength of the bonds in a protein and those it makes with substrates, and another that describes the outside the box effects of such changes on the organism. It is my understanding that polyphen, in part, compares sequences to known databases and scores them based on similarity to give an indication if a given sequence is moving towards them at the level of secondary structure.

1 Like

It is not possible in the way you are thinking, or in the way Behe needs it to be. However, as I already pointed out,

And about this:

This is true, only in part.

1 Like

Yes, that’s why I said it seems like he cherry picked. I’m not overly concerned with Behe’s reputation, as I had developed the idea of his “1st adaptive rule of evolution” long before he ever published it.

1 Like