Behe Meets the Peaceful Science Forum

Some comments on this piece by @NLENTS.
I quote him in italics and my comments are in bold characters.
Presumably, polar bear APOB must be highly optimized for handling the high cholesterol burden.
This is a reasonable inference
Somehow, Behe draws the opposite conclusion
No, not really. What he said in DD is that the mutations in polar bear APOB were likely to be damaging. This is not incompatible with the notion that polar bear APOB must be highly optimized for handling the high cholesterol burden. Don’t forget that APOB is multifonctional, and, as a result, you can have damaging mutations affecting a particular function but not the others. In this situation, it is entirely possible that the resulting damaged APOB protein could be better for coping with high cholesterol burden.
This is puzzling because he notes that humans and mice with mutations that diminish the function of APOB are more prone to suffer heart disease
Agreed, this is kind of puzzling, for, as far as I can tell, it doesn’t seem to strengthen his case.

But polar bears do not get cholesterol-driven heart disease, so the logical inference to draw from these data is that polar bear APOB is enhanced , not diminished, by the mutations.

Given the functional complexity of APOB and our lack of knowledge regarding lipid metabolism in polar bear, I don’t think this conclusion is warranted.

Thus, the location of the APOB mutations provides more support that enhancement, not diminishment, has occurred.
In the case of a multifunctional protein such as APOB, this is untrue.

This program (polyphen-2) can predict whether a given mutation diminishes the function of a protein based on how closely it corresponds to human genetic variation that is known to cause pathology
As far as I can tell, this is incorrect, for PolyPhen-2 also relies on structural features to assess the possible diminishing effect of mutations

But is this even a question? I’m sure each of us can provide several examples off the top of their head


Okay, but if it’s better at coping with the high cholesterol burden, then it wouldn’t be damaged, would it? At this point, we would now have to quibble about what the word damaged means in the context of a multifunctional protein. But considering that the only function being discussed in DD is cholesterol clearance, that’s real rhetorical gymnastics just to rescue his unfounded claim.

Of course. If there other functions that are relevant to this discussion here, Behe didn’t mention them.

Maybe, maybe not. But there is simply NO reason to jump to his conclusions. It’s murky for sure, but the point is that his position, on balance, is the LESS supported inference and therefore offers no support for his thesis.

I disagree. Notice I didn’t say it was definitive, but that it was “more support.” And I defend that. Functions are often mostly housed in specific domains in proteins. That’s why you can often rescue essential functions by providing just small parts of a protein.

Yes, but these are inferred based on the functional/clinical data. The structural inferences (based on the nature of the amino acid change) are applied after the program identifies critical regions using the training it received with the clinical data. @swamidass - isn’t that correct?


It is the question Behe is asking.


IMO many scientists are confused about this.

So from an educator’s perspective, what do you think is the best vehicle for explaining this to laypeople and undergraduates? I can think of 3 candidates:

  1. Why endangered species need us to facilitate outbreeding
  2. Complex inherited diseases in which the variants have extremely low penetrance, as in my field of inherited cardiomyopathies
  3. Incomplete lineage sorting

This is not true, for the passage on the polar bear APOB example in DD, which comprises only six little sentences, doesn’t mention cholesterol clearance at all.

I disagree. Behe has at least the PolyPhen-2 data as well as a very compelling mice study to support his view. His opponents have no experimental data to support the opposite view that the mutations found in polar bear APOB might be constructive at the biochemical level. This is well explained by Behe himself in the piece below.

And for the record, here is the mice study that supports Behe’s case.

How can Behe claim that the mutations were not constructive using PolyPhen when PolyPhen does not even have “constructive” (or anything similar) as a possible output?

Using an inaccurate algorithm trained on human sequence to make grand claims about polar bear protein function just isn’t going to cut it. Behe needs more data to support the claims he is making.


PolyPhen is a machine learning algorithm that translates information (such as structural inferences) to functional effect (from clinical data). So yes, it does take structural inferences to into account, but only to identify what has no effect vs. what has negative effect. It doesn’t have a category for a positive effect, because the clinical data doesn’t annotate things that way.

For this reason, it is common (and correct) to conceptually interpret its output as either “positive or negative effect” vs. “no effect at all.”

The fact of the matter is that we have no computational way of objectively distinguishing between positive vs. negative effect. A mutation that increases the hardiness of cancer is positive for the cancer, but negative for the patient. All algorithms can really do is predict how likely it was that there was a change in function, and (with the right data) whether or not those changes are likely to benefit the cancer or impact the patient.

Polyphen is not operating at that level though. All it is really doing is predicting if there is a change in function. It is an unequivocal error to construe its predictions as determining the biochemical “degradation” of proteins. That is, flat out, not what it is doing.


It’s surprising that I have to say this again, but he does not “have” the PolyPhen data. As much as he tried to manipulate it to say what he wants it to, it doesn’t.

And the mice study isn’t evidence for his position, either. The APOB+/- mice have reduced levels of circulating cholesterol. Polar bears have elevated circulating cholesterol. Quoting from the paper, “Total plasma cholesterol levels were also reduced in the heterozygotes, albeit to a lesser extent ('20%). Because the mouse carries most of its plasma cholesterol on non-apo-B-containing particles(i.e.,HDL), this result was also not unexpected.” To be clear, I’m not saying that its evidence that APOB is enhanced in polar bears but it’s certainly not evidence for his position, either.

But we haven’t made that claim, other than to say that it’s possible. We’re just claiming that there isn’t evidence that it’s diminished.

But the point that is bigger than all others is that none of these variants are fixed in the polar bear population.


And the PolyPhen web page even explicitly states that it is based on human data to be used for human variants. The authors of the paper should not have used it; Behe capitalized on their error to misinform the public.

PolyPhen-2 is an automatic tool for prediction of possible impact of an amino acid substitution on the structure and function of a human protein. This prediction is based on a number of features comprising the sequence, phylogenetic and structural information characterizing the substitution.


Gil, this point needs emphasis. There are no PolyPhen-2 data. There are only its interpretations that Behe is further misrepresenting.

Please don’t misrepresent interpretations as data.


I think this is incorrect. Take a mutation within the active site of an enzyme. Don’t you think we can predict whether it will have a negative effect on the enzyme activity?
Now, regarding the issue of whether some of the mutations found in polar bear APOB lead to loss of molecular function, why not using VIPUR, a method specifically designed for this purpose? Do you think it may help resolve the issue? If yes, maybe someone here with the required skills/tools could perform the analysis.

I know that it is correct and that you are wrong, from the evidence.

Absolutely not!!!

Gil, we’ve literally done this to two different myosins. The most radical side-chain difference possible without changing charge, changing a residue that basically touches the ATP substrate in the active site of an enormously complex enzyme. More than one substitution tested worked, BTW.

Do you ever address any of the many points on which your assumptions are shown to be dead wrong? Do you not realize that you are inadvertently and repeatedly testing the predictions of your ID hypothesis (which you often misrepresent as facts, although you were more modest on this one) and falsifying it?


This tool was published 5 years ago and seems not to be widely used. (I haven’t checked Scopus but citation in PubMed Central has been sparse.) Looking at the citations in PMC, I see two very recent publications about similar tools. If someone wanted to do the “analysis” of this data set, they would have multiple up-to-date peer-reviewed tools at their disposal. Links below, starting with the final peer-reviewed version of the VIPUR paper. See also this previous discussion of the tools (in 2019) and their effectiveness.


And in thinking this, you would be wrong.

Can you even determine if a “negative effect” means increasing or decreasing the enzymes activity? I can’t. There isn’t enough information to tell.


So one approach that Behe could have done is use changes in the predictions of structural stability as a proxy for “degradation”. This is closer to what he means by devolution, and the predictors for protein stability have become reasonably good so as to answer important questions (in aggregate at least).

For example, we could ask, for a random mutation, how likely it is to increase or decrease protein stability. We could ask the same about adaptive mutations.

I haven’t done the experiment, but I expect that:

  1. the adaptive mutations would either be the same distribution as the random mutations, or maybe slightly skewed.

  2. Random mutations in mammalian proteins would be almost as likely to increase stability as they are to decrease stability. If I’m wrong here, I still don’t think that increasing protein stability would be a vanishingly unlikely event.

  3. We would see differences in these distributions between mammals and bacteria, possibly with bacterial proteins more optimized for stability and therefore more likely to be destabilized by a mutation.

So those are some hypothesis to test that could be tested with computational predictions. These are far closer to testing Behe’s devolution theory than Polyphen too.


I am not a biologist, but as an intelligent systems guy I think about dynamic and structural forces all the time. So I wonder whether decreasing stability should be characterized as “degradation,” or whether it is simply one of the axes in a trade-off such that it might be helpful in some situations and harmful in others.

Analogizing from household tools, I find that I need some that are stable (a hammer), some that are very flexible (a plumbing snake), and some that are in-between (a socket wrench). If it is also the case that proteins of various stability can be helpful in various processes (and this is a big if, I’m not a biologist so I am asking a question), why would a decrease in stability necessarily be characterized as a degradation?




True, but a key component of Behe’s definition of degradation is that it is “biochemical” and often points the opposite direction from “function.” He also does address protein stability as something very unlikely to evolve, vis a vis Doug Axe.

So the discussion of “harmful” or not misses the point. Behe is arguing that the beneficial mutations are biochemically degrading, because the vast majority of mutations are biochemically degrading. So we need a definition of biochemical degradation that is independent of beneficial/harmful.

Protein stability fits that bill more precisely than anything Behe has committed too, but frankly he never tractably defines biochemical degradation in the first place!

1 Like

Okay, but if my household tools were slowly evolving, I would think that a trend toward greater stability in my plumbing snake would result in a degradation of its capability.

So I’m putting the question to the biological utility of “decreasing stability” == “degradation.” And if the definition is not helpful for a broad analysis, why should anyone commend the definition to Behe?

Again, I am not a biologist, so I am asking this question so I can learn.


EDIT: It seems to me that if the best answer is that there is no universally applicable definition of protein degradation, then that should be the response to Behe. “You cannot make a universal definition protein degradation because the desirable attributes of a protein vary from context to context” would make a powerful counterpoint to Behe’s analysis–if it’s true. Which I as a non-biologist do not know.


You are right that stability is not by itself good or bad re “function.” If by “stability” we mean “length of time the protein hangs around” then this is clearly not strictly related to function: in fact, some proteins are rapidly degraded as part of their role in a signaling system. (The classic example here is the ubiquitous Wnt signaling system, in which one key protein is constantly made and then rapidly destroyed. Its “stability” is normally low, and this is a feature not a bug.) Some proteins (especially aggregates) become toxic by being hyperstable.

This, BTW, is why the term “degradation” is so confusing (perhaps intentionally so). In a lab studying or using biochemistry, degradation is a pretty specific term that refers to an active process inside cells. Whole fascinating systems are devoted to the destruction of proteins and other cell components, and this is called degradation, indeed often targeted degradation. Cf. the ubiquitin system, in which proteins are “tagged for degradation.” Now walk down the hall of the biology building and listen to some evolutionary biologists or systems biologists chatting. They will talk about “degradation” in the way you seem to use it: degraded performance, maybe degraded fitness, etc. In our particular conversation here, this has significant implications for keeping the conversation coherent and even intelligible.