Behe: Responding to the Polar Bear's Fat

From the paper Behe is citing:

Even with weak evidence from predictive computer algorithms, only around half are damaging. Behe seems to gloss over this fact. Fig. 4 from that paper is worth checking out. The largest proportion of mutations have a 0-20% probability of being damaging.

He also assigns a “possibly” damaging to 10 out of about 25 of these amino acid substitutions. Well if he don’t actually know, then sure it’s “possible” they’re damaging. But how likely is that? If it’s 0.0000000000003% probable then it’s still “possible”.

This mere possibility of being damaging, if it is to be an argument in favor of his main thesis, should be significantly more likely than not.

His argument, as presented, is extremely weak.

In the essay Coyne points to, Nathan and I talk about data and comments apart from the computer predictions that make this very statement.

First, let’s look at the function of apoB:

apoB serves as a tag on cholesterol particles in the blood stream which bind to receptors on cells. Once bound, the cholesterol can be transported into the cell.

So what is happening in polar bears? If apoB were “damaged” then it would have a harder time filling its role as a tag for cholesterol transport. Instead, apoB appears to be doing a better job in its function.

From what I can see, polar bear apoB isn’t broken or damaged. Instead, it is working better than it did before.

1 Like

This is a list of seventeen genes described in Liu et al. as being affected by missense mutations, and their classification according to computer predictions. I have made four groupings here so that readers can see how the computer predictions sort out: those genes whose associated misssense mutations are all benign, those bearing either benign or “possibly damaging” mutations, those with at least one “probably damaging” mutation, and those with missense changes are predicted to give differing outcomes (and may have only benign or possibly damaging mutations). Readers are invited to look into, and comment on, these, and especially on the prospects that these may be degraded or destroyed in the polar bear.


Possibly damaging:

Probably damaging:

Conflicting prediction:

1 Like

Behe posts this at the end of the article:

It’s a list of all the variants predicted to be “possibly damaging” or “probably damaging” by Liu et al. The problem is that he doesn’t inform his audience that this table is only a subset of Liu et al’s table S7. He calls it the “relevant information” from table S7.

The actual table looks more like this:

Are we supposed to believe that Behe did that little switcharoo without thinking that a large fraction of his lay readers would just see a big list of entries saying “damaging” and come away with the impression that these were the majority of variants, or even all of them?


Thank you, thank you, thank you.


But y’see, it lost the ability to tag cholesterol particles at the level it was capable of previously! /s

NOTE: “damaging” in this table is a misnomer, and can mean “good change of function” or "bad change of function"

Table S7
Gene Protein position Ancestral AA Polar bear AA HDivPred HDivProb HVarPred HVarProb
ABCC6 655 Q H probably damaging 0.966 possibly damaging 0.547
AIM1 530 T S benign 0.035 benign 0.018
AIM1 632 I S benign 0.002 benign 0.003
AIM1 821 N K possibly damaging 0.651 benign 0.122
APOB 716 N K possibly damaging 0.801 benign 0.265
APOB 749 D E possibly damaging 0.946 possibly damaging 0.807
APOB 2623 D N probably damaging 1 probably damaging 0.989
APOB 3920 T P possibly damaging 0.482 benign 0.088
APOB 4418 L H probably damaging 0.999 probably damaging 0.915
ARID5B 457 N K benign 0.126 benign 0.02
ARID5B 775 H Q benign 0.006 benign 0.002
ARID5B 787 S R benign 0.412 benign 0.133
ARID5B 875 H Q probably damaging 0.996 possibly damaging 0.731
COL5A3 149 R S probably damaging 0.999 probably damaging 0.993
COL5A3 694 K N probably damaging 0.998 probably damaging 0.991
COL5A3 963 P T benign 0.262 benign 0.249
COL5A3 1117 D E possibly damaging 0.955 possibly damaging 0.804
CUL7 508 D N possibly damaging 0.736 benign 0.159
CUL7 709 D E benign 0.002 benign 0.007
CUL7 1477 N K probably damaging 1 probably damaging 0.999
EHD3 269 K N benign 0.255 benign 0.197
IPO4 362 R W probably damaging 0.977 benign 0.339
LAMC3 791 D E probably damaging 0.992 probably damaging 0.971
LYST 1046 D Y possibly damaging 0.947 benign 0.272
LYST 1084 A S benign 0.045 benign 0.015
LYST 1140 F L benign 0 benign 0
LYST 1770 L F benign 0.005 benign 0.007
LYST 2672 N K benign 0.032 benign 0.021
LYST 2978 R S probably damaging 0.957 benign 0.445
LYST 3784 Q H probably damaging 0.998 probably damaging 0.983
OR5D14 57 F L benign 0 benign 0
OR5D14 306 L F benign 0.014 benign 0.021
OR8B8 48 L V probably damaging 0.999 probably damaging 0.998
POLR1A 413 K N possibly damaging 0.93 possibly damaging 0.496
POLR1A 1317 L V benign 0.001 benign 0.006
POLR1A 1467 G V benign 0.322 benign 0.233
SH3PXD2B 548 Q H benign 0.003 benign 0.002
TTN 995 S I possibly damaging 0.755 benign 0.277
TTN 5869 N I benign 0.326 benign 0.139
TTN 8261 E D benign 0.288 benign 0.109
TTN 8337 I M benign 0.188 benign 0.095
TTN 26365 E D probably damaging 0.992 possibly damaging 0.715
VCL 296 E D probably damaging 0.979 probably damaging 0.982
VCL 600 S R probably damaging 0.996 probably damaging 0.99
XIRP1 868 S R benign 0.145 benign 0.021
XIRP1 1266 G R benign 0.044 benign 0.019
XIRP1 1378 T N possibly damaging 0.718 benign 0.126
XIRP1 1592 S R benign 0.005 benign 0.005

For those keeping track, this is now a total of 8 response articles on ENV alone to the Science review and its “spin-offs” in the last week. They can’t be very busy.

1 Like

Hehe. So true. When you assume something is designed from the get go then any change, no matter what it is, will be viewed as damaging. This seems to be a bias that runs through most ID literature.

1 Like

This gets to a key point. Polar Bears NEED an APOB to function more effectively than it does for brown bears. So yes, the APOB (in a perverse sense) loses a lower level of function to get an a higher level of function. In what backwards sense is this “damaging”?

I’d point out too that these predictors are just predictors, and they are commonly used in cancer where, by convention, even gain of function mutations are called “damaging.” For this reason, they assume that a FUNCTIONAL mutation = DAMAGING mutation. This, however, is not correct in this case. The purportedly “damaging” mutations could very well be the gain of function mutations. Seee, for example, the score outputs for HDivPred and HVarPred:

Prediction outcome can be one of probably damaging, possibly damaging, or benign. “

Notice, it cannot be “improved function.” The reason why is that the program assumes that change of function mutations are damaging, and this is likely the case in cancer (in that they cause cancer). It is not true, however, in the case we are applying it.


I was just thinking about this, since I looked at the paper describing the program that makes the predictions:

It was trained on human disease-associated variants, but as far as I can tell there was no effort made to distinguish between “loss of function” and “gain of function” variants, for example.

1 Like

Exactly. All the program can do is flag mutations as “altering things from ‘normal’ somehow”. It is just by accident it is labeled “damaging.” This is well known in the field. The software, for example, does not have a category for “beneficial.”

I think it also takes into account aa side chain chemistry and possible structural changes. But yeah, it is based on human disease variants.

1 Like

@nlents It is worse than that! @swamidass is “some other guy” as it is his picture he uses. :rofl:

From the article in the opening post:

It appears that this is exactly what Behe did with the polar bear apoB data. He exaggerates the functional predictions, and takes those qualified predictions and makes them into absolutes. Hoisting and petards comes to mind.

1 Like

Exactly - this is a remarkably concise description of 95% of ID rhetoric.

This is an interesting sentence to parse, which Behe appears to be defending:

Since few experiments can be done with grumpy polar bears, they analyzed the changes by computer. They determined that the mutations were very likely to be damaging — that is, likely to degrade or destroy the function of the protein that the gene codes for.

Take the first sentence:

Since few experiments can be done with grumpy polar bears, they analyzed the changes by computer.

Well, this is true. Take the second sentence:

They determined that the mutations were very likely to be damaging — that is, likely to degrade or destroy the function of the protein that the gene codes for.

Who is the “They”? Clearly the researchers that published the polar bear article. What is clear from @Art’s quotes in the article (and @NLENTS), is that “They” did not determine them to be damaging. The table of discussion is one piece of data that was combined with others to conclude otherwise.

A much more troubling problem is that he does not even link to your article @Art (and some guy named @NLENTS). That is a problem.