Gil's testable ID hypothesis

[Rumraket (Mikkel R.)]

@Giltil

Okay. Let’s try to see what is the amount of information (I) produced by a stem cell that differentiates into a naive B cell producing an antibody with a given specificity S.
We have (I)=-log2( P ), with P being the probability that the stem cell gives rise to a naive B cell producing an antibody with specificity S.
Given that estimates of the theoretical diversity of the human naive B cell repertoire range from 10^12 to 10^18, it can be said that P range from 10^-12 to 10^-18 and, as a result, that (I) range from 40 to 60 bits. Note that this range for (I) is an overestimation, for I assume here that there exists only a single antibody with specificity S in the whole repertoire, which in most cases is obviously false.

I am glad to see that you are acknowledging that recombination even by itself, is capable of generating information, even if the sequence fragments pre-exist in the genome of the B-cell.

But I find it rather odd that you’re suddenly interested in calculating just (I) instead of FI, functional information. You must by now recognize that you can’t actually calculate the FI for antibodies, because you have no idea what the total number of molecules of that length that meet the minimal threshold for function is. And for reasons explained in that post of mine you still have not acknowledged, Gpuccio’s method is incapable of providing any sensible estimate.

@Giltil

Note also that (I) should not be confused with FI, for FI is related to the improbability to generate the antibody randomly from an unrelated state

No, calculating FI has nothing to do with how the sequences is generated. It is simply a measure of the fraction of sequences that meet the minimum threshold for function, out of all possible sequences of equal length.

What you’re saying doesn’t make any sense, because that would render FI completely useless as a measure of functional information in evolving sequences, since all sequences evolve from other sequences(either by recombination, or mutation). So they are related to them, they can’t evolve “from an unrelated state”. They can change so much that they become so dissimilar we no longer can see any sign of them having been related through, but that isn’t necessary for there to be FI in the molecule.

How the sequences that meets the threshold, come into existence, whether by design, or by completely scrambling and overwriting some existing sequence, or even by a single substitution mutation, is irrelevant. In either case, the amount of FI exhibited by that sequence would be the same because the ratio of sequences that fall inside, to those outside the minimum threshold, would remain the same.

@Giltil

Bottom line: the maximum amount of information that the process of VDJ recombination is able to produce range from 40 to 60 bits, far, far away from the 500 bits required to draw a design inference. IOW, neither this process nor the process of somatic hypermutation invalidate ID.

No, the real problem is that you don’t actually know how much FI is exhibited by any molecules and you have no way of even estimating it(Gpuccio’s was a commendable but ultimately flawed attempt to get at that). You could try to do it empirically, but for proteins of even small size (30-mers) the total sequence space becomes so large it becomes practically impossible to do thoroughly. So the design inference doesn’t actually work because you couldn’t ever be in a position to say with sufficient certainty that the molecule in question exhibits 500 bits or more.

In my view, the issue is not whether VDJ recombination and somatic hypermutation generates “high FI”(I don’t know what counts as high, is it 500 bits?). I don’t know whether it does that, the problem is that the design inference doesn’t work as long as it’s based on FI.

[Rumraket (Mikkel R.)]

@Mercer

My point is that all of those mechanisms, not just VDJ recombination , produce high FI.

But what is “high FI”? And how would you know the number of sequences that meet the minimum threshold for function?

I have to be consistent here and ask you too. If anyone is going to claim that some sequence or system exhibits some amount of FI, they need to explain how they know that.

[Faizal_Ali]

How long have we been discussing FI, and we still have no definition of what FI is even. It’s long past time for the creationists to just admit they’ve been pulling a fast one, and be done with it.

[Rumraket (Mikkel R.)]

@Faizal_Ali

How long have we been discussing FI, and we still have no definition of what FI is even.

We do have a definition, Hazen et al. came up with one in their 2007 PNAS paper “Functional information and the emergence of biocomplexity”.

It’s just that it’s really not useful in practice, because those damned thresholds are almost impossible to estimate for any sufficiently long sequence. Gpuccio tried to do it with some mangled attempt at using sequence conservation over deep time, but for reasons I keep referring back to, it doesn’t work.

[Dan_Eastwood]

please hold relies while i split this topic.

[Mercer (John Mercer)]

@colewd

Even if this is an enzyme we could estimate the total amount of fitness peaks through experiment.

So, let’s get back to the point and test Gil’s hypothesis:

@Giltil

« the genetic information coding for the V regions of the immunoglobulin receptors within B cells largely preexists in the genome of the stem cells that develop into B cells »

Note that Gil moved the goalposts a lot. Let’s test the hypothesis that he initially offered:

@Giltil

“V(D)J recombination doesn’t produce high FI because most of the genetic information necessary for implementing the function preexist in the stem cells that develop into B cell”.

Again, Gil, there is (at least to me) an extremely obvious experiment that allows us to test that hypothesis. What happens if you get rid of most of that germline information?

We don’t even have to mess around with the sloppiness of @gpuccio’s hypothesis because we can immune measure function itself, not by any sequence proxy.

Why do you keep running away from any engagement on this matter?

What are you afraid of?

If you can’t see what the experiment is, why aren’t you curious?

[Mercer (John Mercer)]

@Rumraket

But what is “high FI”?

In the experiment Gil clearly does not want to consider, FI is simply information that produces high actual function, in this case, immune function. I’m not proposing to do this by BLASTing, but by directly and quantitatively measuring immune function.

@Rumraket

I have to be consistent here and ask you too.

Fair enough.

@Rumraket

If anyone is going to claim that some sequence or system exhibits some amount of FI, they need to explain how they know that.

By assaying function. Gil doesn’t seem to be interested in anything but moving the goalposts. Why is that?

[Mercer (John Mercer)]

@Giltil

Okay. Let’s try to see what is the amount of information (I) produced by a stem cell that differentiates into a naive B cell producing an antibody with a given specificity S.
We have (I)=-log2( P ), with P being the probability that the stem cell gives rise to a naive B cell producing an antibody with specificity S.
Given that estimates of the theoretical diversity of the human naive B cell repertoire range from…

…

@Giltil

IOW, neither this process nor the process of somatic hypermutation invalidate ID.

Gil, since you’ve trice, and very explicitly, limited all of your consideration to naive B cells, I don’t see how you are addressing somatic hypermutation in any way.

I don’t see why the naive repertoire is relevant if we are concerned with functional information, as the antibodies produced after maturation are much more functional than the ones displayed by naive B cells.

Can you explain this huge discrepancy? At what point in B cell differentiation are you thinking that somatic hypermutation occurs, exactly?

[Giltil (Gilbert Thill)]

@Rumraket

Note also that (I) should not be confused with FI, for FI is related to the improbability to generate the antibody randomly from an unrelated state

No, calculating FI has nothing to do with how the sequences is generated. It is simply a measure of the fraction of sequences that meet the minimum threshold for function, out of all possible sequences of equal length.

Are you aware that when you say that « FI is a measure of the fraction of sequences that meet the minimum threshold for function, out of all possible sequences of equal length », you are in fact saying the same thing that what I am saying, ie, that « FI is related to the improbability to generate the antibody randomly from an unrelated state »?

[Rumraket (Mikkel R.)]

@Giltil

Are you aware that when you say that « FI is a measure of the fraction of sequences that meet the minimum threshold for function, out of all possible sequences of equal length », you are in fact saying the same thing that what I am saying, ie, that « FI is related to the improbability to generate the antibody randomly from an unrelated state »?

You can calculate that probability from the FI, sure. But the sequence doesn’t have to come into existence by being generated “randomly from an unrelated state” to have X amount of FI. As I was saying, how the protein comes into existence is irrelevant to how much FI it has.

[Mercer (John Mercer)]

@Giltil

Are you aware that when you say that « FI is a measure of the fraction of sequences that meet the minimum threshold for function, out of all possible sequences of equal length », you are in fact saying the same thing that what I am saying, ie, that « FI is related to the improbability to generate the antibody randomly from an unrelated state »?

Are YOU aware that testing your hypothesis:

@Giltil

“V(D)J recombination doesn’t produce high FI because most of the genetic information necessary for implementing the function preexist in the stem cells that develop into B cell”.

is very clean, because it doesn’t require using proxies for function? That we can test the extent to which immune function has been implemented directly?

I just don’t understand why you aren’t either

1. pointing out how there is no experiment that can test your hypothesis,
2. asking how this experiment tests your hypothesis,
3. expressing interest if you don’t see how this experiment tests your hypothesis.

You keep claiming that you know things by misrepresenting testable hypotheses as mere claims, but you show zero curiosity about their veracity.

It’s as though you have absolutely no faith in your claims and prefer to frantically move the goalposts around in avoidance of information!

[colewd (Bill Cole)]

@Jordan

Of course I’d ask the same of @Giltil and other ID folks that want to contribute. I think we need a good-faith attempt to work with the science from both sides.

I have not read the hole thread but it appears Gil has proposed a practical way to test the hypothesis. Opponents can always nit pic that the test is inadequate but what Gil proposed will improve our understanding and it should be straight forward.

[Rumraket (Mikkel R.)]

@colewd

but it appears Gil has proposed a practical way to test the hypothesis.

From what post does it appear that way? It seems to me it’s John Mercer who is the one proposing to test it.

[colewd (Bill Cole)]

@Rumraket

From what post does it appear that way? It seems to me it’s John Mercer who is the one proposing to test it.

They both are proposing tests. One way to figure out if Gils claim is correct is to find if the exons sequence pre exists in the genome. This should be straight forward if the the sequence data is available.

[Rumraket (Mikkel R.)]

@colewd

They both are proposing tests

Where’s the post where Gil proposes an empirical test?

@colewd

One way to figure out if Gils claim is correct is to find if the exons sequence pre exists in the genome.

No, that wouldn’t be testing Gil’s claim because that would be to confuse sequence with functional information. Not all sequences have functional information, regardless of how similar they are to functional sequences. It is whether it meets the minimal threshold for function that matters.

[colewd (Bill Cole)]

@Rumraket

No, that wouldn’t be testing Gil’s claim because that would be to confuse sequence with functional information. Not all sequences have functional information, regardless of how similar they are to functional sequences. It is whether it meets the minimal threshold for function that matters.

If they can be combined to kill an antibody then by definition they contain functional information.

[Rumraket (Mikkel R.)]

@colewd

If they can be combined to kill an antibody then by definition they contain functional information.

When they have been combined such that they have a function yes. Not individually. If the individual fragments to not have the function of interest, they do not have FI. Hence, recombination (and mutation) generates the functional information.

[colewd (Bill Cole)]

@Rumraket

When they have been combined such that they have a function yes. Not individually. If the individual fragments to not have the function of interest, they do not have FI. Hence, recombination (and mutation) generates the functional information.

I disagree. You are nit picking. If the exon sequence is mission critical that is where most the information is.

[Rumraket (Mikkel R.)]

@colewd

I disagree.

Your agreement is not required, go read Hazen et al 2007 again. The concept is defined there, sequences without function do not have FI.

Here let me quote the relevant piece:

In this formulation, functional information increases with degree
of function, from zero for no function (or minimum function) to
a maximum value corresponding to the number of bits necessary
and sufficient to specify completely any configuration of that
system.

There’s more:

Functional information is defined only in the context of a specific function x. For example, the functional information of a ribozyme may be greater than zero with respect to its ability to catalyze one specific reaction but will be zero with respect to many other reactions. Functional information therefore depends on both the system and on the specific function under consideration. Furthermore, if no configuration of a system is able to accomplish a specific function x [i.e., M(Ex) = 0], then the functional information corresponding to that function is undefined, no matter how structurally intricate or information-rich the arrangement of its agents.