I know.
I still think you’re putting too much into it. Keep it simple, even for the spectators.
1 Like
But that’s your idea. There are billions of different ATP synthase variants, are they all “optimal solutions”?
I can go find some species, find the sequence for it’s ATP synthase protein subunits, compare them to some other arbitrarily picked species’ corresponding ATP synthase protein subunit, see that they are considerably divergent(down to 30% sequence similarity or even less, for proteins 500 amino acids long or more), and then I have to contend with the observation that it supposedly obtains thermodynamically “perfect” conversion of chemical into mechanical energy to within the accuracy of measurement of the experiment you referenced(given that all measurements have some degree of uncertainty, I am also surprised the authors would even have the gall to claim they have shown it is perfect).
Disregarding for a moment that I find that to be a fantastic claim (which deserves it’s own thread to be honest, I’m not sure the authors of that paper even know what they’re saying, as it would imply the machine imparts no kinetic energy to surrounding water molecules when it spins and bumps into them, which it of course does), it still leaves YOU with having to explain how it is possible for so many different ATP synthases to retain perfect energy conversion despite considerable differences in amino acid sequence of it’s constituent proteins.
I ask again, do you think they all are perfect? Or only that one of them? If only that one of them is perfect, and I can show by phylogenetic tree how it evolved incrementally from a non-perfect ancestor, what is the problem for me again? Have I not then shown how the global optimum was found by a gradual process?
1 Like
Oh and on a related note, you’re using the efficiency of thermodynamic conversion as a substitute for the “fitness” of the ATP synthase machine, which it is not obvious to me that it is.
I would not need persuasion that it is one component of it’s fitness, but it has to do more than not produce waste in the interconversion of chemical and mechanical energy. It also has to actually work at sufficient speed to produce ATP fast enough to drive competitive growth and division, resist denaturation and degradation to some degree, etc. etc.
All of these components contribute to it’s fitness, and so it is actually likely(or at least possible) there is some compromise between these different aspects of it’s function where it’s performance is sub-optimal, and that a higher-fitness alternative exist somewhere in sequence space, where it is 99.999999% efficient, but 15% more resistant to degradation(thus needs to be replaced significantly less often, saving energy in that way instead), and 1.05% faster(say).
If you do not know the efficiency, stability, speed, and accuracy of other ATP synthases, and how these components affect it’s fitness, then how can you claim to know how they are distributed in sequence space, and that this one you referenced is the globally optimal solution in terms of all attributes that contributes to it’s fitness?
Not only the volume knob, but all of the knobs on Nigel’s amp go to 11. 
2 Likes
FI being defined as the -log2 of the ratio of the target space to the sequence space, it follows that for a FI of 700 bits, this ratio is equal to 1 / 5,26.10^210, which, I am sorry to tell you, means exactly that the sequence space is 5,26.10^210 bigger than the target space. This is not a matter of hypothesis, it is a matter of definition. Do you understand this?
I understand that you are channelling Nigel Tufnel.
You are making zero attempt to measure either the target space or the sequence space, Gil. The hypothesis that you clearly are afraid to test is that this calculation is an estimate of that.
This silly “FI” calculation is Nigel’s knob. You simply aren’t measuring the thing you claim you’re measuring.
You are trying to sneak in a hypothesis as fact without having the courage or faith to test it in the real world.
2 Likes
I know that. Nigel’s knob doesn’t really tell us what the volume is.
I agree. The calculation is ridiculous. However, for the basis of others reading, it seems far simpler and more intuitive to address the falsehood of the underlying assumption: that there is a correlation between conservation and functional information.
There just isn’t.
2 Likes
Agreed, which I why I spent time detailing how conservation is explained by natural selection having driven some sequence to a local optimum, and why the evolutionary sampling of sequence space is extremely biased, and why the number of sequences sampled for even a relatively short protein is only an infinitesimal fraction of that total space.
- On what a local optimum is in the fitness landscape analogy.
- On how the sampling of sequence space around a particular protein is highly biased.
- On how only a tiny fraction of the total space is sampled.
And as we saw above, handwaves in the direction of “statistics of sample size”, or the ruggedness of the fitness landscape, do not constitute meaningful rebuttals to these issues either.
3 Likes
The system is in a highly biased condition without a doubt. How did it get there?
This is not a big issue for statistical sampling.
How do you know this is what you are observing? Are you saying that other optimum sequences exist given the current configuration of myosin?
All those questions have been addressed in our numerous lengthy exchanges, there’s zero reason to rehash it all over again.
1 Like
If you and John think assertion is a reasonable standard I agree, however I don’t think you have clarified the difference between system bias and sample bias which is an important issue.
I don’t think you have established “islands of function” in proteins that exist in a system like muscle proteins.
I don’t think John has established that amino acid sequence conservation and functional information are not correlated other than asserting it is not correlated. When he did his test it failed miserably. Myosin has 6000 more bits of FI than alpha actin per blast results.
I will agree to disagree at this point and just want to establish there are still open questions.
YOU haven’t established that FI and sequence conservation are correlated. On top of that, there is every reason to think that they aren’t correlated since sequence conservation is due to evolutionary mechanisms that find local optima.
That does raise a legitimate question in it’s own right, about exactly when you’d infer a local optimum has been reached, as opposed to a locus just having a very low rate of change due to most mutations being strongly deleterious?
For a protein like alpha actin 1 that appears to have remained completely unchanged for half a billion years or more, I think it’s safe to say there are no equal-or-greater-fitness variants within hamming distance 1. And probably none within a distance of 2 either, possibly even 3?
You could imagine that there are some extremely unlikely mutational variants of alpha actin 1, say a version with >5 simultaneous substitutions, that technically have comparable fitness. It’s just that such a specific multi-substitution that could in principle bring it across a deep fitness valley between two peaks, would be very unlikely to occur and also fix, if it’s not also even more fit, as opposed to merely neutral.
But there are some sequences that appear to “only” be evolving very slowly, which implies there are still a few rare neutral or adaptive mutations among all possible single or double mutants, instead of having got truly stuck at that one single best one in it’s local sequence neighborhood.
There are also contingencies, or frozen function. This happens when multiple other proteins bind to an arbitrary position on the protein of interest. Once a protein-protein binding interaction occurs then you will have strong selection against changes in both proteins. However, the sequence itself is somewhat arbitrary.
It’s been a while since I looked at actin, so I did a quick Google search. It appears that there are many deeply ancestral proteins that interact with actin, so I’m thinking that the optimum for actin involves more than just the ability to polymerize. When there are so many other proteins interacting with binding sites on actin then strong conservation is expected.
Yes certainly those would be the constraints that define the local optimum, that make it be a local optimum. It is not that the particular sequence is the only one that could do the job, but to find another local optimum would require multiple simultaneous changes to all the interacting polymers.
All good points. I’ve also been thinking about what internal constraints operate on actin filaments as they slide between myosin filaments, and I have this intuition that the protein needs to retain structural integrity and prevent break or misfolding as this molecular sliding is bound to generate a lot of friction, which would also imply selection on probably every single residue.
But like you said, this particular combination is probably arbitrary, and if we go back far enough on the tree of life and into other clades, we’d probably see how another such local optimum in homologoues of alpha-actin-1 had evolved.
And many other reasons, all of which are on display in the cardiac sarcomere.
But Bill, despite bringing up the sarcomere himself, can’t seem to muster up the courage to answer a simple question:
How many MYH7 (protein) sequence variants have been found in healthy humans?
It’s as though Bill knows he won’t like the answer…
Unlikely. It’s more about historical contingencies as mentioned by Taq above:
Presumably you don’t get external interactions on the inside of a folded globular protein, which then raises the question of why residues in there appear not to undergo any change for many hundreds of millions of years? That implies they are under some other constraint than protein-protein interactions. It could “just” be intramolecular constraints then. That they epistatically interact with other residues both on the inside and on the surface of the protein.
1 Like
I suspect that the answer is both, probably synergistic.
Of course, the remarkable thing that Bill can’t explain is the enormous polymorphism of the cardiac myosins, despite their functional complexity. Do you think that there is any way to get Bill to see this with open eyes and an open mind?
Nope.
2 Likes