When I read the title of this thread, I honestly thought that was where they were going: That they were going to use this to double down on their belief that COVID-19 was created in a lab.
The spike protein also consists of two subunits, each of which consist of multiple domains. Some of which are even known to be able to fold autonomously, without being part of the entire spike protein(in particular the receptor binding domain alone can do this, and it’s just about 200 amino acids long). The RBD alone has about 15 mutations total.
Whichever way you look at it, this protein should be pretty much half-way if not entirely dead if Miller’s confused assertions reflected reality.
Now the real problem with his assertion is his misreadings of the literature he cites. Particularly what he takes the Tawfik study to imply.
It is true that the average effect of mutation without selection has a destabilizing effect on protein structure, and it also makes sense that the magnitude of the destabilizing effect of each individual mutation would diminish as the protein grows larger (because, basically, the larger the protein the more internal surface area can “stick” together so any individual amino acid would be a smaller part of the structure). But Brian Miller takes this to imply one protein can’t evolve to acquire a new function (in the video he basically asserts that the flagellin protein of the flagellum couldn’t evolve from a protein with a different function) by accumulating mutations, because he appears to not understand that purifying selection can maintain structural stability of a protein against destabilizing mutations, such that you can continue to accumulate mutations well beyond the naive threshold you would reach without selection. His idea seems to be that if two proteins are too different from each other in sequence, then there is no way to get from one functional protein to another even if they’re as similar as 90%, because by the time 10% of the sequence has changed the protein should be completely nonfunctional.
That is absurd.
We know of examples of proteins that diverge in sequence basically all the way from 100% to 15%(or even below in some cases) sequence similarity, while retaining a functional and stable structure. And we have divergent superfamilies where we know of variants that diverge at large portions of the way, at increments on the level of 1-2% at a time. 100%-98%-96%-94%(…)10%-8%-5%-3%. Incidentally there was a good example in table 2 of the paper I referenced in this thread(and again here).
Mostly dead implies a significant reduction in viral fitness given the crucial function of this protein to the lifecycle of the virus. There’s no evidence it has suffered any reduction in fitness. On the contrary. Omicron is rising in frequency, and rapidly outcompeting the delta variant. It would be very odd if it was accomplishing this with a protein that falls apart much more frequently due to destabilizing mutations, than it’s delta-variant competitors.
Now the broader context of Brian Miller’s talk is his attempt to try to debunk the reality that proteins can evolve to change functions, such that one protein performing some function A can be mutated and co-opted to perform a different function B (in the video talk he casts this in the context of the flagellum filament protein flagellin evolving from a protein with a different function than being a filament). He is quite clearly arguing that this can’t happen, since apparently he appears to think the accumulation of mutations in some pre-flagellin protein would render it nonfunctional long before it acquires the ability to function as a flagellin.
So I’m just left wondering here. When is the spike protein going to fully fall apart? Why do we have so much evidence of many other proteins diverging, and having already diverged, from their common ancestors, well beyond the sorts of limitations(be that 3-10 total mutations, or ~10% of total protein sequence) Brian Miller is misreading his references to imply?
One of the side effects of a loss of stability is the specificity in binding is reduced, allowing weak binding to more targets. For obvious reasons, that does not necessarily result in lower fitness. The problem here is that fitness defined according to evolutionary theory is not a helpful model in understanding protein structure or function.
Let’s look at the specific data here, Ben, rather than traffic in generalities.
From the original variant to omicron we observe 30+ mutations. If I am understanding you correctly, you would expect the omicron variant to be less stable and to bind less specifically to the ACE-2 receptor. Is this a fair and accurate reading of your statements?
What other human receptor than ACE2 do you have evidence Omicron spike interacts with, to a greater extend than previous variants? Sounds to me like you’re making things up here.
Rumraket hits on this several comments down, saying Brian Miller “appears to not understand that purifying selection can maintain structural stability of a protein against destabilizing mutations.”
Yes, that is clearly implied by Miller trying to argue against a case of a protein evolving from being able to perform one function, into another similar protein that carries out a different function. He’s being quite clear this is what he is trying to accomplish with this argument. He’s not just trying to argue that to discover a new functional protein de novo is very unlikely, he’s also trying to argue against the idea that, for example the flagellum’s filament protein - flagellin - could have evolved from another protein that was less than 90% similar to it.
By Miller using research on protein stability, where mutations were allowed to accumulate without selection to see how many it would take on average before the protein destabilized, to argue against proteins being able to accumulate mutations over time, he really does appear not to understand the rule of purifying selection in these sorts of random walks in protein sequence space.
Yes, when selection is not included, then after relatively few mutations the protein is pretty much guaranteed to destabilize. Hence mutations without selection can’t just accumulate willy-nilly and indefinitely, and you would, without selection, expect the protein to become nonfunctional long before a new function was found (supposing, just for the sake of argument, that 10% of the sequence has to change to find a new function).
The whole point here is that the very same research indicates that when purifying selection is included, the protein really can continue to accumulate mutations well beyond those 5-10 amino acid substitutions, or 10% of sequence length, or whatever “limit” Brian Miller thinks there is. I make this same point in this post.
But why does Rumraket think Miller is talking about a case of purifying selection? In Rumraket’s second source, Miller even says he’s talking about a case of weak selection: “After only a few random mutations (1-2) under weak selection , around a third of subsequent changes to a protein completely disable it.”
This fellow is confused. I am not arguing that the strength of selection changes the proportion of mutations that are destabilizing versus stabilizing. I am arguing that Miller is ignoring selection when he argues against the long-term accumulation of mutations in protein sequences. When Miller is trying to cast doubt on the possibility that, for example, the flagellum’s filament protein flagellin could have evolved incrementally from another protein that is less than 90% similar, by appealing to research done trying to establish how many mutations it takes, on average, to destabilize a protein.
That research shows it takes only a few without selection (because on average more mutations are destabilizing than stabilizing), but when selection is included, the protein actually doesn’t destabilize because now the destabilizing mutations are either selected against(those mutants have lower fitness because their protein falls apart more often), or being compensated for by the fixation of stabilizing mutations.
This is an absurd misunderstanding of the waiting time problem. The waiting time problem refers to when you need two simultaneous, specific mutations before a fitness gain is realized.
No, it just refers to two(or more) specific mutations. The whole problem with the waiting time argument (I’m now explaining what is wrong with the waiting time problem-argument) is that it asks us to pick out particular combinations of mutations (two or more) and then asks us to calculate the odds that that specific “target” set of mutations would occur. And since it would take a lot longer to wait for two specific mutations, rather than to wait for just any two mutations, any time we find that two sequences are different from each other by some number of mutations we could argue they couldn’t possibly have evolved from each other (or from a common ancestor) in the available time because our calculation shows it would take an inordinate amount of time before that set specific set of differences to have occurred. I have a post that explains the problem with the waiting time-type of arguments here:
No reason to go over the rest of this fellow’s post because he doesn’t seem to know what I’m referring to. Yes, creationists (such as John Sanford) really have argued that to get some specific set of mutations takes a really long time on average, and therefore evolution is false because we find that two things A and B have that many differences, and that is regardless of whether those mutations each confer a selective advantage or not.
And he’s actually right. The math generally checks out. It really would take a really long time to wait for that specific set of 2 or more mutations. It’s just that those aren’t the only possible set there was, it was just the one that occurred.
It doesn’t apply when:
The mutations can happen one at a time with a gain in fitness at each step.
Flat out wrong. It still applies. The math still shows that if you need to wait for two specific mutations, even if each of them carries a fitness benefit, it will take a loooong time for those two specific mutations. And if you pick out three specific mutations, it gets even worse. And so on and so forth. The whole argument here is about specific target sequences vs any sequence that works.
If anyone thinks this is all just “creationist nonesense,” here’s Larry Moran saying the same thing:
“The probability of any single mutation occurring is equal to the mutation rare, which is about 10-10. The probability of an additional specific mutation occurring is also 10-10. The combined probability of any two specific mutations occurring is 10-20… Let’s say that three specific mutations are required to change from a cluster of two needles to a cluster of five needles. One hundred million years ago you could calculate that the probability of three specific mutations is about 10-30. It’s highly improbable, just like the specific bridge hand. When such a triple mutation arises we recognize that it was only one of millions and millions of possible evolutionary outcomes.”
Uhm, yes, exactly. That’s my whole goddamned point. That it makes no sense to pick out specific sets of mutations after the fact and declare them to be a “target sequence” that couldn’t possibly evolve because it would take too long(nor does it make sense to calculate the mean time to establishment of some X number, as that treats them as a target too), exactly because it is just one possible set that evolved out of many others that could. It was never a “target” to begin with. It’s just a contingent outcome of history.