Behe on genes subject to positive selection

I thought I would un-bury this.

Am I reading this correctly? Are all genes subjected positive selection expected to be broken if Behe is correct?


It would be great if others can chime in here.
@Art: You picked up the part of Behe’s argument that intrigued me the most in his article. If it’s testable, perhaps you can explain how and challenge Behe/other ID scientists to do the same.

Not all, just most. Overwhelmingly the most. Yes, it is testable.

What is striking about this is that Polar Bear data is a great way to test his hypothesis. It shows his reasoning. We see that his reasoning is based on a misunderstanding of Polyphen 2. It is possible the reason why he does not want to concede the point here is that doing so would evaporate most of his case.

Notably, he doesn’t seem to realize that breaking genes is self limiting. If, for example, you have 1000 genes that you can break (that are not essential), then you can only break 1000 genes. The fewer genes like this you have, the less breaking you will expect. In contrast, the processes that generate new FCT (as he calls them) are not limited this way. They will go on at about the same speed.

“Devolution” is, by necessity, self-limiting. Constructive evolution is not.


What would the number of “genes” available to break depend on? Or is the situation too complex to make such a call?

Depends on the system.

What If we take polar bears as an example. How many genes would the polar bear genome have that are available for “devolution”?
Or is it something impossible to say?

What we know is that there was a handful of genes selected.

It does not appear that ANY of them had obviously degrading mutations (blunting or breaking), unless I missed something in my read (and maybe I did). That is significant, because obviously breaking mutations are possible (e.g. truncations), but I don’t think they were observed. (that being said, truncations sometimes increase function)

Most these proteins have mutations that are predicted to be change of function (though the algorithm, as a quirk, calls them “damaging”). We do not know if any of these mutations are breaking/blunting or if they are improving function. We do not have objective criteria for determining this any ways, even if we tested them in an experiment.

1 Like

Generally, one could identify genes subject to positive selection and characterize them for function with the computer (best in cases where there are many candidates) or one-by-one. An example of the latter approach is in this paper. (Note that this paper also is relevant to the matter of promiscuous enzymes, and that the wild-type product of the positively-selected gene is functional.)

I guess 99/100 or 999/1000 is just about “all”. But “almost all” is probably better.


I can see why you would say this.
If a gene Impacts more than one function, then the situation can be complicated.
Also things can get messy if “function” is “that which is selected for”… in some cases blindness is a function that is selected for.
If function is viewed as a “job done” by a gene, then changes that reduce efficiency can be viewed as “damage” can’t it?

That loophole is important for Behe. If we produce an example of a gain of function, he will claim it is the exception. The fact that he relied on computer simulation is interesting and substantially weakens his position. With a better program, one that doesn’t call them “damaging”, he won’t have the fig leaf any more.

The bigger issue is this:

Even directly testing things, he could (ignoring exaptation) claim that the gene was broken in the original context (brown bears) to work better in the new context (polar bears). Of course, the same argument would work reverse, if brown bears evolved from polar bears. It is not based on any biochemical measure of “function”.

This makes his argument very weak, especially if he ever descends to arguing in that way.

How would “function” be defined for the computer.
Is it how efficiently a particular “job” is done…
Or based on how much a change improves fitness?

Or am asking a nonsense question :slight_smile:?

1 Like

Exactly. ANY change in function can be called “damage” in a mendacious way. Let us say, for example, that ApoB has increased function. Someone, in backwards way, could claim that it had “lossed its lower level of function,” which is technically true but mendacious. Evolution requires change. Change means that things change. Change is, by definition, loss of one state to gain another.

If you focus on the “loss” and ignore the “gain”, yes, you can tell an absurd story that this is a problem for evolution. That misses the point that every loss, in this way, is also a gain.

By some sane measure, we could build (or find) a predictor for “change of function.” It would not tell us which direction the change was, but such a thing is plausible.

1 Like

Can you give an example to make things easier to understand.

1 Like

Here are examples “change of function” prediction:

We can check to be sure, but these algorithms don’t have the quirk of calling everything “damaging.” We would have to look at the precise techniques used to ensure they are appropriate for this task. We could then put the mutations to predict what they do. The predictions should be interpreted as “change of function”, not loss or gain, which (as I explained) is trivially true. Every change of function is BOTH a gain and a loss of function, in one sense or another.

It’s a great question. Behe focuses on the immediate - a specific sort of enzymatic activity or biochemical parameter - so we can too. Computer predictions can do some of this (PolyPhen 2 is an example) but, as we have seen on this board, they give at best tentative results. Direct assays are better, but can get laborious, especially if we want to put to the test the 99/100 or 999/1000 numbers Behe tosses out.

As Joshua is noting, there is lots of wiggle room that could allow Behe to imply damage, even if there really isn’t. But if Behe’s thesis is can only be saved by resorting to this sort of hair-splitting, then it isn’t worth much.

1 Like

@Ashwin_s, did you see the presentation I gave today?

I show how “damaging” mutations can be used to explain every step in constructive evolution (the Muller two step) in order to create irreducibly complexity. According to Behe, damaging mutations are easy, so evolving irreducibly complexity is predicted, by Darwin Devolves’ thesis, to be very easy to evolve. And it is. Behe never gets a chance to see this, because he doesn’t engage with the implications of his thesis in combination with exaptation and neutral theory.

I should add that it would be easy to identify clearly broken genes with the computer. So there is a chance to affirm Behe’s thesis in an unequivocal way, for each positively-selected gene, if dramatic loss-of-function is involved.


And, of course, we don’t see many truncations (any?), which does not support Behe’s thesis.

I’d also point out that at least one truncation event in human brain evolution increased function, verified by experiment, and this seems to be a key step in human evolution (@bjmiller).

1 Like

I guess the entire problem depends on what one takes as frame of reference. I can think of two options to summarize.

  1. A particular environment is the frame of reference-
    In this case, anything that helps to survive better would be an improvement/gain of function. In such a scenario, the idea of “damage” is meaningless.

  2. Efficiency to carry out a particular job (like a biochemical activity, or particular morphology) is considered.
    In this case, one can talk about damage in terms of loss of efficiency doing that particular job…

So Behe’s Argument would be that Evolution predominantly with loss of function kind of changes which turn out to be advantageous in a particular environment.
Is that correct?


But look at #2 closely. We EXPECT there to be change of function mutations, where one biochemical activity is lost and another is gained. This is what builds up complexity in biology. That definition does not take exaptation into account.

1 Like