Below is Gpuccio’s answer to your question, which I endorse : How long must the time window be so that sequence homology may be considered a good estimator of FI? I would say at least 200 – 400 million years. Better if 400. Why? Because that’s more or less the time window that is usually associated with “saturation” of synonymous sites, IOWs with the more or less complete loss of any detectable homology in neutral sequences.
Wait, hold up… Are you saying that us not knowing what the function is indicates that there may in fact not be one? Or that it is fair to assume there is none until after we could at least vaguely tell that there might be one? That’s… unexpected.
Given that nobody has measured the Functional Information of the microorganisms contained in my drop of water, your claim that my claim is “wrong” is unsubstantiated.
That is my point. Unless and until FI has actually been measured, everything claimed about it, including all of Gpuccio’s claims, and all your defenses of his claims, are nothing but pure unsubstantiated speculation, with no more weight than my zillions of bits in a water drop claim.
@Giltil: what evidence do you have that 400MY sequence homology is a better estimator of FI than this week’s lottery number (or some other random number)?
Why do you endorse it? You’ve failed to answer all the explanations for why it doesn’t work. How is this not just blind contrarianism in the end, then? A sort of loyalty test perhaps, or based in some sort of theological commitment? Gpuccio’s defense of ID is your religion?
Because high level of conservation through deep time indicates high constraints which indicates high FI. Take two proteins P1 and P2 of the same length showing respectively 85% and 55% conservation through deep time. Isn’t the case that It could reasonably be concluded that P1 has higher FI than P2?
Your opinion, which I don’t share. Let’s agree to disagree.
The “reasonably” part is what we are asking. What links the observation with the inference outside of mere word games? Because, of course, if we define “FI-richness” as conservedness, then sure enough it follows, but only in virtue of silly word substitutions. What, if there is such, is the unique feature of FI that is not contained in full or in part by conservedness alone (i.e. what is the novelty, what makes the whole discourse around FI non-trivial), how do we measure that, and what measurements has anyone obtained, what data was gathered from whence one could assert a correlation between the two?
The only philosophical commitment befitting of undertaking scientific inquiry is one to intellectual honesty with oneself and one’s peers. No colours, no names, no creeds must be above the data nor a substitute for such. If you do not know a thing, it is a matter of self-respect to admit as much to yourself and to go on to either fix it or learn to live with it. If you struggle to defend what you think you know against your peers’ criticism - let alone the criticism of those who are not your peers, but masters of the land you merely tread into - then it is time to reconsider if you do in fact know what you think you know, and if their disagreement may in part be rooted in them knowing something you do not. And if the data itself apprently flies in the face of what you think you know, then insisting that you do, to yourself or to others, is a lie, unbecoming of someone with aspirations to getting taken seriously at all, when they speak on scientific matters.
High constraint isn’t a measure of FI at all. It is completely incapable of telling you the actual amount of FI (as explained before and you’ve never answered).
At best, all it says is FI for a more conserved protein is likely to be higher than FI for a less conserved protein. That’s not an actual measure of FI, it is at best an estimation of a relational property (A > B, or A < B).
Yes. But that’s not a measure of A, nor of B. Telling me Peter is taller than Frank tells me nothing about their actual heights.
Even if it was weakly an indication of FI (it isn’t, as explained before numerous times), that still wouldn’t allow you to infer ID. As also explained before and you, remarkably, appeared to concede. So why are you now back to making all these same assertions?
What is the major malfunction at work here? I’m going to need some sort of explanation rather than your recurring disappearance from threads such as these. What happens when you move on? You just… forget it all?
No, not good enough. Respond to the criticisms you have been offered here before. You can’t just wave your hands and move on. That shows you have a deep issue with how you approach reason, logic, debate, and evidence. Some sort of weird philosophical game, or worse, some sort of psychological issue is at work.
No, but mine don’t beg the question against any particular “world view” being advanced. Including yours. I do not have some sort of axiomatic commitment to naturalism. At best I have some basic assumptions about the approximate reliability of our senses, and that I am at least capable of some degree of reason, and that they inform me about an external world. I can’t prove that they do in any ultimate sense, but that’s basically as far as that goes. Notice how none of that entails, nor even implies, that it should not be possible for me to discover that something has been intelligently designed, for example. I don’t have “philosophical commitments” that should be considered problematic in our disputes about design and evolution. But it does actually appear you do. You seem unable to actually let go of ideas that can’t be defended with actual evidence and reasons. You have this problem, not me.
Not unless lack of constraint indicates low FI. Remember that you’re trying to compare FI between sequences. If low constraint doesn’t indicate low FI, you can’t say that the highly constrained sequence has higher FI than the unconstrained one. As people have been telling you, you can keep at most one hypothesis from 1) Gpuccio’s measure of Fi works for his purposes and 2) most of your genome is functional. I say “at most” because neither is defensible.
Another important issue is the relationship of conservation to FI, which has not been addressed. I’m not even sure you have a definition of FI that you’re willing to commit to. It’s certainly not relevant to Szostak’s definition.
That’s not an explanation of your reasoning, though, is it? It’s just parroting.
So far, so good, and in your own words! But you are missing a qualifier for “constraints.”
Not even close. That’s a vapid assertion, contradicted by what we know empirically. For example, conservation of the actins represents STRUCTURAL constraints. That doesn’t mean that actin contains more functional information.
So, in your own words, why would this apply to orthologs and not to homologs? Members of protein families (like the myosins) have overlapping functions and have diverged over your parroted definition of deep time. All of @gpuccio’s reasons still apply.
Do you not realize how obvious it is that you haven’t thought about this in any depth?
I think we have to be careful about which FI we are talking about.
We could be talking about the intuitive concept of ‘functional information’ (which it seems is what you are discussing). This, as you suggest, leads us to a bunch of intuitions, but no hard conclusions.
We could be talking about “Functional Information”, as defined by Szostak. This has, AFAIK, never been measured for any real-world situation. This lack of measurement means that we have no idea what is a good proxy for it, and what it is a good proxy for. This is in the same way that the fact that nobody has (seen let alone) measured a unicorn means that we have no idea whether a Falabella or a Shire Horse is a better proxy for one. For this reason, we also have no way of knowing the extent that our intuitions about ‘functional information’ hold true for Szostak’s definition. This is the definition that @Giltil claims as “the definition Gpuccio uses”, so it was this that I was talking about above.
We could even be conceivably talking about Durston et al.'s measure of functional information, Functional Sequence Complexity. This measure appears to show no more relationship to Szostak’s defintion than the fact that they are both, in different ways, trying to quantify the same intuitive concept. It is this measure that shows some weak relationship to protein conservation, but this is merely that “it can be observed that a high conservation value usually corresponded to a high measured FSC value” but that “[t]he measurement is also affected by the number of amino acids observed which could be different for different sites.”
@Giltil’s defense of Gpuccio appears to be a hot mess of fallacious equivocation between the three, combined with an a priori assumption that high functional information must be a sign of Design.
If we know something about the biology and biochemistry and biophysics, we don’t if we are contrasting myosin with actin. All of those definitions are fuzzy, but myosin obviously has much more than actin. No intuition is necessary. That was my point.
As an outsider with rather few intuitions on the matter, I do not find this obvious at all. Having more complexity that serves function (which is what I intuit is meant by “functional complexity”) is one thing, but unless we either define that this is what having functional information means or is implied by, or have some working model of functional information that links it with functional complexity in this way, I wouldn’t necessarily infer that it therefore must have more functional information. The claim doesn’t overtly violate any gut feelings I might have on the subject, to be fair, but a “therefore” between one and the other would raise questions in me and I don’t feel particularly uncharitable because of that.
What is functional information outside of what is fully covered by other, better established quantities, how do we measure it, and what data do we have showing a correlation between them strong enough to hinge inferences like that upon?
We really can’t, because we are leaving digital for analog.
The best metaphor I can offer now is that we can agree that a locomotive has more of whatever functional parameter we want to define/calculate than the track it runs on, but that’s an exaggeration because to be an accurate metaphor, the track would have to self-assemble.
Among the worst part of it all is that Gpuccio’s argument directly depends on how much variation is sampled in the alignment. That’s a giant problem right there. If technically 10 out of the 20 different possible amino acid states are allowed at some position in the protein, but only 5 of these are among the variation known from species that have had the gene sequenced, that’s massively going to impact the FI you calculate as you go through each aligned site in the protein and there are strictly allowed variants for each position not actually present in existing variation.
This idea that we would expect all the allowed variation to actually be present in the form of sequences from each different species is just ridiculous. Particularly when we consider that some states might just have lower fitness and therefore be less likely to be present, yet would still strictly represent functional states, and therefore affect the probability of evolving the sequence.
The method is therefore extremely vulnerable to taxon sampling, and completely ignores the effect of fitness. It treats each site as black or white, with variants not seen in the taxon sampling assumed to be nonfunctional, which will inflate the FI. Besides the fact that it also assumes that the taxon sampling yields an exhaustive sampling of allowed variants.
And then there’s the fact that saturation of synonymous sites tells you nothing about how exhaustive the sampling has been of multi-residue mutations for amino acids.
Amazingly I’ve explained all this on this same website now four years ago. I got crickets back, like we get crickets now. In the latest FI thread I also linked an article co-authored by Swamidass where they prove with simulations that the sampled variations you can expect to be present in some taxon sampling used for your alignment, to “fantastically understimate” the real number of functional sequences.
It’s stupid. It’s just stupid. I cannot get over that there are people who buy it. They must be having trouble actually understanding what is going on. But, we’ve been over all this before so many times. No response.