Hunter: Finally, the Details of How Proteins Evolve

Conversation
24

That doesn’t make sense. Testing hypotheses about past events giving rise to observed data is always about finding patterns more consistent with the hypothesis, and less consistent with others. If you want to call that “arguing from evolution” then it isn’t clear to me that “arguing from evolution” is any sort of a problem.

Yes, what is the problem here? That description makes good logical sense. Repetitive microsattelite DNA is the sort of thing known to be prone to duplication, which is further corroborated by the fact that similar loci in closely related species show variable length of these repeats.

I’m sure you understand the concept of a metaphor.

In the press release Cheng admits that AFGP evolution “occurred as a result of a series of seemingly improbable, serendipitous events.” For “not just any random DNA sequence can produce a viable protein.”

I’m left wondering so what? You’re not explaining how that is a problem, or even that it is a problem?

Furthermore, in addition to the gene itself, “several other serendipitous events occurred.” The DNA was “edited in just the right way,” and “somehow, the gene also obtained the proper control sequence that would allow the new gene to be transcribed into RNA.”

And?

Serendipity/Just-So is prevalent in evolutionary thinking. It is a monumental problem.

So you say, but it’s not clear to me what the problem even is. You seem to have said what amounts to “this particular chain of events looks now, after the fact, as an unlikely combination”. But wouldn’t that be true for ANY conjunction of serial stochastic events?

4 Likes
25

We see the same in geology and astronomy of course. Observed data or phenomena are explained by invoking certain theories about forces in the present also operating in the past. The distribution of elements in the Earth’s crust and mantle, the geological features on the surface, the prevalence of Earthquakes in particular areas and so on, are exlained by invoking various physical theories about plate tectonics.

Looking at the distribution of matter in the solar system into planets, asteroids, and comets, their locations in relation to each other and their observed trajectories are explained as a consequence of physical forces like gravity having operated for billions of years into shaping the observed patterns. This can be extended to basically the entire cosmos. Globular clusters, trails of dust and gas, galaxies and entire clusters of them are also explained by invoking these forces.

The past gave rise to the present conditions through the physical forces acting on the “initial conditions” in effect at the time.

IDcreationists seem to have a hard time fathoming that this exact same elementary concept is how evolutionary biologists explain presently observed patterns in genetics and anatomy by invoking observed mechanisms such as mutation, drift, and natural selection having operated in the past given the conditions at the time. There really is no in principle difference here.

4 Likes
26

How would one test the scenario in question?

Here’s the problem (the same problem occurs throughout this entire story, actually). Why include the adverb “fortuitously” in this sub-hypothesis?

“A putative translocation event in the last common ancestor of AFGP-bearing gadids moved the hitherto unexpressed AFGP precursor to a new genomic location that fortuitously contained a TATA motif thereby enabled transcription.”

Well, OK. Serial stochastic chains occur. That’s what Markov processes are.

But “fortuitously,” and similar language that occurs in the press release, which George noted – e.g., “serendipitous” – suggests that we are being asked to accept the occurrence of a long string of very low probability events, which the authors themselves acknowledge are unlikely, and flag as such with appropriate modifiers.

Testability?

2 Likes
27

What a strange question.

How would one test a scenario for how some mountain range, or a particular mountain formed?

The basic scientific principle about explaining observed patterns that arose in the past essentially take this form: If X happened we should expect to find Y. Finding or not finding Y is the test. That’s the only way to do it.

Here’s the problem (the same problem occurs throughout this entire story, actually). Why include the adverb “fortuitously” in this sub-hypothesis?

How is that a problem at all?

“A putative translocation event in the last common ancestor of AFGP-bearing gadids moved the hitherto unexpressed AFGP precursor to a new genomic location that fortuitously contained a TATA motif thereby enabled transcription.”

That’s what happened to the citrate transporter gene in the LTEE. A locus was duplicated into another part of the genome and it just so happened to end up under control of a promoter that was active under aerobic conditions, and the translocated gene just so happened to be a citrate transporter that could take up citrate, which just so happened to be present in the growth medium.

Stuff like this happens.

Well, OK. Serial stochastic chains occur. That’s what Markov processes are.

But “fortuitously,” and similar language that occurs in the press release, which George noted – e.g., “serendipitous” – suggests that we are being asked to accept the occurrence of a long string of very low probability events, which the authors themselves acknowledge are unlikely, and flag as such with appropriate modifiers.

What’s the probability of your particular set of mutations? What’s the probability of your parent’s particular set of mutations? What’s the probability of their parents sets of mutations? Compound those probabilities back in time 10.000 generations.

Testability?

Are we to re-create planetary orbits too, and regrow forests, rivers, glaciers, mountains, and tectonic plates? How did the world happen up so arranged as to give rise to the exact structure we recognize as the Mt Everest? To think of all the ways the combination of erosion and plate tectonics had to play out in minute detail to explain the precise location of every atom that makes up the Mt Everest. That event very likely only happened once in the entire history of the observable universe, and likely won’t repeat again before it undergoes heat-death.

How do we “test” the theory of the origin of the Mt Everest? Try to think of how you would have to create, for example, a computer simulation and let erosional forces and plate tectonics produce an exact replica of the Mt Everest down to the atomic scale.

Some of you seem to be demanding something absurd from science about explaining past events, and I doubt you hold that same level of evidence for any other beliefs you have. Are you a creationist? How would you test creationist claims about how this gene originated?

3 Likes
28

I’ll take that as the long form of “this scenario can’t be tested – it’s a narrative, not a testable hypothesis.”

Which was George’s original point, in his blog post.

Some narratives are true; others are not. Very hard to say in this case, because we are banking on the very lucky role(s) of several unrelated dice. The point George was making is this scenario gives us no independent grounds for thinking it is true, other than the alternative is creationism. Since the alternative is magic (for most biologists), low standards of testability apply. This has been a standing weakness of evolutionary theory since Darwin.

1 Like
30

Lotteries are designed to be unpredictable. That’s how they work.

Are you saying that evolutionary narratives are designed to be untestable? Or just that such untestability happens to be a logical consequence of their narrative structure?

31

We can and have tested the mechanisms which produce the results. We don’t have to recreate the specific results. Seriously, do you actually use this dumb rhetorical argument anywhere?

32

I will disagree. The experiment - transfer, duplication, activation of genes - is ongoing, and can be studied computationally and in real time. Specifically, in plants, there ha been a large-scale movement of genes from organelles to nuclear genomes. This is an ancient process, but one that is occurring today, and one that can be studied in real time. The basic bottom line from all of this is that the activation of new genes isn’t (in geological or evolutionary timescales) a particularly improbable event. This includes the recruitment or de novo origination of organellar localization information.

To be sure, more work needs to be done. But the notions that these processes are beyond the reach of direct experimentation, and that they are fantastically improbable, are wrong.

4 Likes
33

Evolution also doesn’t have any pre-specified results. That’s how it works.

How would a geologist test the idea plate tectonic collisions caused the Himalayas to rise? We can measure plate tectonic movement (in mm per year), calculate the forces involved, find additional physical evidence the process did indeed cause the mountains to rise over the last 50 MY or so. Because we can’t actually recreate the same mountain ourselves is that evidence space aliens built Everest and K2?

I realize Creationism relies on empty rhetoric in lieu of scientific evidence but wow, you’re really hitting the bottom of the barrel with the “evolution is untestable” nonsense.

34

You are restating my point in different words. The entire narrative – what you call “the specific results” – is untestable.

Unhappiness with the overwhelmingly narrative logical structure of evolutionary explanation, in the late 1970s, led many younger investigators out of neo-Darwinism. Joel Cracraft, for instance, complained loudly:

The critical factor in the scientific study of any phenomenon is that explanatory hypotheses should be susceptible to criticism, that explanations should be evaluated by empirical tests of some sort. As far as I can see, statements of the type that “phenotype x is an adaptation, evolved via the agency of natural selection” are thoroughly untestable. The necessary data needed to refute such an assertion cannot be gathered, and we are more or less forced to accept it as an article of faith rather than as a scientific statement.

J. Cracraft, “The Use of Functional and Adaptive Criteria in Phylogenetic Systematics,” American Zoologist 21 (1981):21-36; p. 32. [Nota bene: Cracraft told me a couple of years ago, in a conversation at the American Museum in NYC, that he is less confident about this criticism of neo-D today.]

Now, maybe Art is right – testability is still available – but that’s not what folks earlier in this thread were saying. We don’t need no stinkin’ tests.

36

“Testing” is most certainly possible. This is a good place to start learning more. Enjoy.

3 Likes
37

Thanks, Art. I have that paper (very interesting work). I don’t think it applies directly to the antifreeze gene origins scenario, however.

Gotta run, thanks to all.

1 Like
38

This body of work speaks directly to the idea that promoter capture is in some way fantastically improbable. Which may be, IMO, an important aspect of antifreeze gene origins.

1 Like
39

Oh my.

Wow.

Wow again. I can’t find the slightest hint of that in anything @Timothy_Horton stated. Am I missing something?

Even though I’ve seen the “evolution is untestable” trope in countless forums, I’m very surprised to see it applied in this particular context.

I heartily commend Dr. Nelson for being so forthright in sharing this anecdote of updated information. Truly. Such willingness to put everything on the table is much appreciated and contributes even more of value to this very interesting discussion. Thank you for that.

No matter how much we may all disagree on various matters, I greatly respect the willingness to engage in peaceful and forthright constructive dialogue.

4 Likes
40

I would say that parts of the narrative are testable and due to the limits of science and the laws of physics, not all of it directly testable. It seems you neglect the possibility that science is limited.

1 Like
41

Which would be ridiculous, because the hypothesis that the gene arose from non-coding DNA is tested by the comparative genetics.

There might be details about the elucidated scenario which are untestable, such as whether some repetitive region was first duplicated and then later mutated, or mutated first and then later duplicated, but that does not mean the more general claim of a protein coding gene evolving de novo from non coding DNA isn’t testable.

The test of that hypothesis was done.

But you seem to mean something else by a test. A test is apparently whether we can show that the differences we see between the different species really arose by mutations. Which we obviously can’t do. For anything. Even experiments that take place right now. How do we really know when we see two DNA sequences that are different from each other, that the difference, which we call a mutation, was caused by some sort of chemistry or physics? How do we know an invisible or occult force didn’t “make it happen”? Well we don’t. I also don’t know that the moon isn’t kept in it’s orbit by such an occult force. It’s all inference.

Suppose we create a giant water tank and freeze the water and have some fish evolve in it, and we DNA sequence these fish every X number of generations and see what looks like adaptive changes take place. How do we know those genetic differences, the mutations, were really caused by some physical or chemical reaction, and weren’t zapped into happening by some occult force or undetectable mythical being?

We can’t. Either way it’s still inference.

Some narratives are true; others are not. Very hard to say in this case, because we are banking on the very lucky role(s) of several unrelated dice. The point George was making is this scenario gives us no independent grounds for thinking it is true, other than the alternative is creationism. Since the alternative is magic (for most biologists), low standards of testability apply. This has been a standing weakness of evolutionary theory since Darwin.

Then propose me your test of creationism. What is the creationist hypothesis here, and how do we test it? We put the fish in the water tank, and then we just wait for the occult creating force to zap the gene into existence?

Your “testability” rhetoric here is a red herring. You cannot provide the level of evidence you are demanding of evolutionary biologists. You can provide nothing at all. At least biologists can make some general (and some times rather detailed) predictions about what closely related species should look like genetically if protein coding genes evolved from non coding DNA.

What does creationism predict? Where do de novo genes come from? Are they magically zapped into being, as if huge insertions not derived from any similar sequence anywhere?

Don’t tell me, the creationist hypothesis is completely identical to the evolutionary one, you ALSO expect a similar region in very similar but this time independently created species, just because… why?

1 Like
42

This is right but what is left is that the design hypothesis which is also an inference is on the same playing field as the “evolution did it” inference.

43

Horrible or great explanatory track records should play a role in our reasoning and in the conclusions we make. “Magic” has a horrible track record. So it can be hard to take that as a serious alternative.

1 Like
44

We got a pretty cool universe from it :slight_smile:

45

When a Markov process is complex, it is basically impossible to re-create the starting conditions. For example, how would you re-create the initial condition of the two continental plates about to collide when the Himalayas were still a plain?

This does not mean scientists can make no progress. The scientific community can observe existing phenomena to derive probability tables for state transitions given certain conditions. Those probabilities can then be rolled into a model that performs a maximum likelihood estimation.

Incredibly, such a model can be highly accurate. Allow me to give an example from natural language processing, a field which faces challenges of stochasticity similar to biology. I built a Naive Bayes model for predicting the geographical origin of test tweets, based on observations of tweets with known geographical origins. The model was able to achieve about 70% accuracy on a test set of tweets, based only on tweet text. The eye-opening aspect of this model was that the probability of the correct origin was typically an almost infinitesimal fraction; it just happened to be the largest probability of all the possible predictions. So the model’s ability to predict accurately looked serendipitous. :slight_smile: But it’s hard to argue with a model that works.

We could regard the LTEE as a kind of test set for the evolutionary model. Against that test set, the model has been highly successful. It hasn’t predicted specific mutations down to the nucleotide, of course; but mechanisms posited by evolution have been observed in the lab, and have produced the kinds of results expected by the theory.

This is why the theory of evolution is considered successful by biologists, and why it is used to explain the changes in DNA that led to the “anti-freeze” de novo gene.

From the perspective of a stochastic process, there were an astronomical number of possible paths from starting conditions millions of years ago, and only one happened. That’s the nature of the universe God created for us and for Himself. From the fish’s perspective the path was serendipitous, and from the mathematician’s perspective the path was highly improbable–just like any other path through natural history.

What makes Zhuang’s paper good science is not that the observed outcome could have been deterministically predicted given the starting conditions a million years ago. No. What makes it good science is that the state transitions described by the paper are commonly observed by biologists.

@Cornelius_Hunter Is your criticism that the most likely path described by Zhuang cannot be ascertained with certainty? If so, let’s stop having jury trials; they are infected by the same uncertainty, regardless of the evidence. Any reconstruction of the path of a Markov process, be it human interactions examined in a jury trial, the causes of the American Civil War, or the appearance of the novel AFGP gene, could easily be labeled a “Just So Story.” What you call a problem for evolution is a problem for every human endeavor. What I am wondering is why you don’t see the problem for any human endeavor other than biological science. Or at least you don’t profess to see it.

Your thoughts, @Cornelius_Hunter?

2 Likes