Common Ancestry and Nested Hierarchy

[misterme987 (Andrew P.)]

Okay, thanks everyone for all of your help on the “Nested Hierarchy and Bootstrap Values” thread. Using different programs than FastML, I was able to get much more reasonable results. For example, here is the tree (both unrooted and rooted) that I got from 16 random ‘taxa’ (visualized using iTOL: Interactive Tree Of Life):

Apparent polytomies, very low bootstrap values (1-30; avg. 12), and other strange things rarely seen in real-life phylogenies abound.

In contrast, here is the tree (both unrooted and outgroup-rooted) produced by 16 ‘taxa’ that share common ‘ancestry’ (also visualized using iTOL: Interactive Tree Of Life):

All relationships are correctly reconstructed, with bootstrap values of 100 to boot (pun not intended).

The same thing was observed in other tests that I ran; random trees always have low bootstrap values and strange topologies, as opposed to trees produced by taxa that share common ancestry, which always have a clear nested hierarchy.

This falsifies the creationist claim that nested hierarchies are merely an artifact of differing levels of similarity caused by ‘common design.’ In any given set of random ‘taxa,’ there will be differing levels of similarity, such that some ‘taxa’ are more similar to each other than they are to the rest of the ‘taxa.’ But, as this test shows, the differing levels of similarity between random ‘taxa’ do not produce a nested hierarchy like the ‘taxa’ which share common ancestry.

Thus, phylogenetics really is strong evidence for common ancestry. Every time phylogenetic analysis produces a nested hierarchy with high bootstrap values, that is evidence that the taxa at the tips of the tree really do share common ancestry. (Whether or not ‘common design’ happened can’t be determined by phylogenetics because common ancestry is compatible with both design and unguided evolution.)

So do real-life taxa produce the same nested hierarchy structure with high bootstrap values? Yes. Just look at any of the thousands of phylogenetic analyses that have been done in the past by evolutionary biologists – in almost every single case, a nested hierarchy is resolved with very high bootstrap values. For example, see this primate ERV phylogeny, or this whole-genome tetrapod phylogeny, or this comprehensive molecular phylogeny of primates.

Just for fun, though, I generated my own phylogeny of primates, using the cytochrome b, cytochrome c, and cytochrome c oxidase subunit I protein sequences from 14 different primate taxa (including Homo sapiens). Here is the resulting tree, both unrooted and rooted, visualized using iTOL: Primates :

There is an obvious nested hierarchy, with comparatively high bootstrap values (46 - 100; avg. 85), and none of the strange topology that dominated the ‘random’ trees. Furthermore, this tree agrees with previous phylogenies that have been done by real evolutionary biologists, recovering all of the major primate clades, as the color-coded tree below shows.

I also generated a phylogeny of 9 tetrapod taxa, using cytochrome b + cytochrome c + hemoglobin α1 + hemoglobin β1, and the result was similar (visualized using iTOL: Tetrapod):

There is again a clear nested hierarchy, with comparatively high bootstrap values (61 - 100; avg. 86), and none of the strange topologies which dominated the ‘random’ trees. Therefore, we can be pretty certain that all tetrapods are indeed related by common ancestry, whether or not ‘common design’ was also involved in their creation.

Strangely, this phylogenetic analysis recovered turtles (represented by Chelydra serpentina) as the sister group of squamates (represented by Anolis carolinensis), but I didn’t think too much of it since the position of turtles in the tetrapod tree is contested.

Another analysis that I did (which can be found at iTOL: Interactive Tree Of Life) recovered squamates (represented by Anolis carolinensis and Pantherophis guttatus) as a single clade, and turtles (Chelydra serpentina) as the sister group of archosaurs, more in line with recent phylogenomic analysis of tetrapods, but this tree had lower bootstrap values (10 - 100; avg. 63). I would trust a whole-genome analysis more than my own simple tree based on just a few proteins, though.

In summary, phylogenetics does provide a real test of common ancestry, as evidenced by a comparison between ‘random’ trees and trees produced by taxa that share common ancestry. The fact that nearly every single phylogenetic analysis thus far has resolved a nested hierarchy with high bootstrap values is nearly incontrovertible evidence for common ancestry.

My own analysis of primate and tetrapod taxa shows that, at least, all tetrapods (including humans) are related by common ancestry. This can probably be extended even further back, all the way to the base of the phylogenetic ‘tree of life,’ since a recent phylogenetic analysis of taxa from all three domains of life shows that Archaea and Eukarya form a single clade (Eukarya bootstrap value 100%; Archaea + Eukarya bootstrap value 100%) and Bacteria forms a single clade (bootstrap value 100%).

@thoughtful, @colewd, anyone else here who disagrees with common ancestry, what do you think about this?

[misterme987 (Andrew P.)]

Anticipating one possible objection: yes, I know that separate ancestry is slightly different than the ‘random’ sequences I generated above. I’m currently working on a simple program (much like my original one that simulates several generations of evolution in a population of related ‘taxa’) that models such a separate ancestry hypothesis, by doing essentially the same thing as the original program but starting from several different ‘ancestral’ sequences rather than just one.

My prediction is that it will look rather like the ‘random’ trees up to the point where common ancestry is actually occurring, and then will look like the ‘common ancestry’ trees. I highly doubt that any kind of strong phylogenetic signal will be recovered at the base of the tree, where the ancestral ‘taxa’ are unrelated. But who knows, I could be proven wrong.

[Joe_Felsenstein]

John H will know more, but I believe that recent evidence is tending to validate your placement of turtles.

[Rumraket (Mikkel R.)]

Well done Andrew!

[Meerkat_SK5]

According to cellular automata, a complete self-replicating automaton must consist of three components: a universal constructor (UC), a (instructional) blueprint and a supervisory unit. These functional components are required to produce successive generations of artificial life, which happens to produce patterns that look like nested hierarchy.

The algorithmic origins of life | Journal of The Royal Society Interface (royalsocietypublishing.org)

[Tree (automata theory) - Wikipedia](Tree (automata theory) - Wikipedia

More importantly, Von neuman’s universal constructor model is an exact representation of common design model. For example, the universal common designer, the common blueprint, and homoplasy/ non-random mutations would be the biological version of Von Neuman’s Universal constructor, blueprint, and supervisory unit.

This means that you are merely assuming those patterns represent common descent or inheritance.
But, those patterns actually represent a common blueprint and mechanism the designer used to front load the reproductive and survival capacity into basic types to adapt to their respective environments (i.e. homoplasy/HGT and non-random mutations/cytosine deamination).

These studies provide support for what I am saying here:

" Functional domains of proteins include sites for phosphorylation, glycosylation, ubiquitinilation, sumoylation and glutathionylation, as well as sites that interact with other proteins. The function of such domains is determined by specific sequences of amino acids that must be coded in the DNA. All functional domains and sites will contribute to sequence identity of homologous genes in separate species. Phylogenetic analyses are therefore in effect a genetic mirage—the result of artificial constraints imposed by analyzing functional domains together with non-random mutations—largely determined by the Physico-chemical properties of the DNA sequence and its environment."

(PDF) Shared mutations: Common descent or common mechanism? (researchgate.net)

“Furthermore, we show that our proposed model accounts for most of the mutations at neutral sites but it is probably the predominant mechanism at positively selected sites. This suggests that evolution does not proceed by simple random processes but is guided by physical properties of the DNA itself and functional constraint of the proteins encoded by the DNA.”

Evolution: are the monkeys’ typewriters rigged? | Royal Society Open Science (royalsocietypublishing.org)

[misterme987 (Andrew P.)]

Okay, I completed my program which models separate ancestry (it can be found here). Essentially, it generates a pre-specified number of random ‘ancestral’ sequences, and then for every ‘generation’ mutates the sequence to produce two daughter sequences. By the end of the simulation, we have many different taxa that fall into multiple, unrelated ‘kinds,’ which is the separate ancestry hypothesis.

Here is the unrooted phylogenetic tree generated by PhyML from 10 different ‘kinds’ of 4 ‘taxa’ each (visualized using iTOL: Interactive Tree Of Life). The taxa are numbered as “kind #_taxon #” (e.g., taxon 4 from kind 6 will be numbered as 6_4):

As you can see, this ‘separate ancestry’ tree has much the same strange topology as the ‘random’ trees (including a polytomy). The main difference is that there exist clusters of ‘taxa’ where each ‘kind’ falls on the tree.

Unfortunately, I was unable to properly view the bootstrap values because of the highly variable branch lengths, so I had to switch over to a different format to view the bootstrap values. Note that the branch lengths on the tree below are not representative of the number of changes along each branch.

As I predicted, the bootstrap values between each ‘kind’ are remarkably low (0-7; avg. 1), while within the ‘kinds’ they are significantly higher, though still not nearly as high as the ‘common ancestry’ tree (which had bootstrap values of 100).

To make sure that these results were not merely a statistical fluke, I ran the ‘separate ancestry’ simulation + phylogenetic analysis several more times. Here is one of the resulting trees (with 5 ‘kinds’ of 4 ‘taxa’ each), both unrooted and midpoint-rooted (visualized using iTOL: Interactive Tree Of Life):

Again, the bootstrap values between ‘kinds’ are extremely low (4-17; avg. 8), while the bootstrap values within ‘kinds’ are higher, though again not 100 as in the ‘common ancestry’ tree. The same thing was observed in all of the other ‘separate ancestry’ trees.

This tells us what to expect from separate ancestry, from a phylogenetic standpoint. If God really did create multiple ‘kinds’ that were not related to one another, then we would expect those ‘kinds’ to have strong phylogenetic signal within them, but between them there would be little to no phylogenetic signal (represented by low bootstrap values).

Contrary to creationist claims, phylogenetics really can distinguish between separate ancestry and common ancestry.

[John_Harshman]

Depends on what placement you mean. I can’t think of a recent analysis that failed to make turtles the sister group of archosaurs.

[thoughtful (Valerie (non-scientist))]

@misterme987

@thoughtful, @colewd, anyone else here who disagrees with common ancestry, what do you think about this?

Just trying to understand what you’re doing first. How do you go about creating a random set of taxa?

[colewd (Bill Cole)]

@misterme987

@thoughtful, @colewd, anyone else here who disagrees with common ancestry, what do you think about this?

Hi Andrew
I like the work you are doing and how you used the resources here to improve your models.

If we assume a consistent molecular clock it appears that the consistent tree pattern is being generated by the emergence of common gene sequences at different times. Unfortunately a consistent clock is not what we are observing between different genes and may be contributing to lower bootstrap values.

From a simple model stand point if two species were either created or were the result of a split from a common ancestor it would appear the tree pattern would group their common genes together assuming a consistent molecular clock and the closest time of origin among the selected species,

In a common design model we may see sequence modifications to similar functioning genes. If you look at the same functioning genes between vertebrates and invertebrates you will see evidence of this.

[Joe_Felsenstein]

@colewd

If we assume a consistent molecular clock it appears that the consistent tree pattern is being generated by the emergence of common gene sequences at different times.

That makes no sense to me. I’m not sure what any of it means.

@colewd

Unfortunately a consistent clock is not what we are observing between different genes and may be contributing to lower bootstrap values.

But the methods used to infer the trees do not assume a molecular clock.

[misterme987 (Andrew P.)]

@colewd

If we assume a consistent molecular clock it appears that the consistent tree pattern is being generated by the emergence of common gene sequences at different times. Unfortunately a consistent clock is not what we are observing between different genes and may be contributing to lower bootstrap values.

No, you’re piggybacking on (and misinterpreting) my comment on the other thread where I said that differing levels of similarity could possibly be explained by separate creation events at different times in history. The entire point here, which I proved in the OP, is that differing levels of similarity aren’t the same as nested hierarchy.

Frankly, I’m not sure how you could misinterpret my comment from before so badly, given that I said (in the exact same comment) that “[Bill’s] hypothesis is still falsified by the fact that these sequences form a nested hierarchy.”

[misterme987 (Andrew P.)]

@thoughtful

Just trying to understand what you’re doing first.

No problem.

@thoughtful

How do you go about creating a random set of taxa?

It’s programmed in Javascript, so for each nucleotide in each random sequence, I use Math.random() to generate a pseudo-random number between 0 and 1. If the number is <0.25, the nucleotide is an “A,” if it is between 0.25 and 0.5, it is a “T,” etc. This creates essentially random sequences of pre-specified length.

I understand that this isn’t exactly ‘common design,’ but there is no way to effectively model ‘common design.’ The point is that many creationists argue that differing levels of similarity produce the same nested hierarchy as common ancestry. For example, this article by CMI tells us that

Since DNA codes for structures and biochemical molecules, we should expect the most similar creatures to have the most similar DNA. Apes and humans are both mammals, with similar shapes, so both have similar DNA. We should expect humans to have more DNA similarities with another mammal like a pig than with a reptile like a rattlesnake. And this is so. Humans are very different from yeast but they have some biochemistry in common, so we should expect human DNA to differ more from yeast DNA than from ape DNA. So the general pattern of similarities need not be explained by common-ancestry (evolution).

From Answers in Genesis:

But what about similarities used for classifying different kinds of animals—like dogs and horses—into categories like mammals or tetrapods? Biblically, we can interpret such similarities as reflections of living things’ shared Creator. This Creator applied certain useful design plans—like four limbs—in multiple applications, to serve similar functions under similar constraints. In fact, like Harry Sanders points out, design often does produce nested hierarchies, even in human-made creations. For example, trucks are more similar to each other than to vans because they were designed for similar purposes. Both trucks and vans, meanwhile, are more similar to each other than to two-wheeled vehicles.

And from ICR:

For example, consider the similarities and differences among major types of transportation vehicles. An Indy racing car has much more in common with a sedan (e.g., four wheels, movement restricted to land, etc.) than with a hovercraft. However, all three vehicles have more in common with one another (e.g., movement restricted to sea or land) than with a helicopter. Thus, a “tree of transportation” could be drawn without much effort by simply observing and classifying the products of design that surround us, and this tree would depict vehicles in a hierarchical pattern

So all of the major creationist organizations are making this claim (and perpetuating this misinformation). It’s not just a straw man.

Since any set of random sequences will necessarily have some sequences that are more similar to each other than they are to the other sequences, the fact that such random sequences don’t produce a nested hierarchy shows that the differing levels of similarity do not, in fact, produce the patterns that we see in real life.

Perhaps there are as-of-yet untested hypotheses for how separate ancestry can produce nested hierarchies. But such hypotheses have not been stated yet, by either separate or common ancestry proponents, and so necessarily the best hypothesis to explain the data is common ancestry. That’s how the scientific method works. Perhaps you have a testable hypothesis for how a separate ancestry model can explain nested hierarchy?

[Meerkat_SK5]

@misterme987

This tells us what to expect from separate ancestry, from a phylogenetic standpoint. If God really did create multiple ‘kinds’ that were not related to one another, then we would expect those ‘kinds’ to have strong phylogenetic signal within them, but between them there would be little to no phylogenetic signal (represented by low bootstrap values).

Contrary to creationist claims, phylogenetics really can distinguish between separate ancestry and common ancestry.

Oh please, all you did was refute a strawman argument. Common design can explain and predict those nested hierarchies WITHOUT violating those patterns. You just choose to assume it was common descent instead.

[misterme987 (Andrew P.)]

@Meerkat_SK5

According to cellular automata, a complete self-replicating automaton must consist of three components: a universal constructor (UC), a (instructional) blueprint and a supervisory unit. These functional components are required to produce successive generations of artificial life, which happens to produce patterns that look like nested hierarchy.

The algorithmic origins of life | Journal of The Royal Society Interface (royalsocietypublishing.org)

I think you are somehow misreading that paper. Walker and Davies seem to only be talking about the origin of life, and a quick search for “nested hierarchy,” “phylogeny,” and related words in that paper turns up nothing. When they are talking about hierarchies, they are not referring to the hierarchical structure of phylogenies, but about the hierarchical structure of gene regulatory networks.

@Meerkat_SK5

Tree (automata theory) - Wikipedia

This is not about biology but about mathematics. So I’m not sure what point you are trying to make. If you’re saying that the designer followed such a pattern, creating an original ‘template’ and modifying it in stages from there, that is basically the same as common ancestry, and I’m not averse to such a hypothesis. I don’t think that whether design happened or not can be inferred from biology.

@Meerkat_SK5

More importantly, Von neuman’s universal constructor model is an exact representation of common design model.

Note that I never claimed to have disproven common design. I merely showed that the nested hierarchical pattern of phylogenies (with high bootstrap values) is evidence of common ancestry over against separate ancestry. Design is compatible with both.

@Meerkat_SK5

This means that you are merely assuming those patterns represent common descent or inheritance.

No, I’m not. That was the entire point of my post, to show that separate ancestry does not produce those patterns, and so it is not an unfounded assumption that these patterns are evidence of common ancestry.

@Meerkat_SK5

" Functional domains of proteins include sites for phosphorylation, glycosylation, ubiquitinilation, sumoylation and glutathionylation, as well as sites that interact with other proteins. The function of such domains is determined by specific sequences of amino acids that must be coded in the DNA. All functional domains and sites will contribute to sequence identity of homologous genes in separate species. Phylogenetic analyses are therefore in effect a genetic mirage—the result of artificial constraints imposed by analyzing functional domains together with non-random mutations—largely determined by the Physico-chemical properties of the DNA sequence and its environment."

(PDF) Shared mutations: Common descent or common mechanism? (researchgate.net)

This article (which is from Journal of Creation) doesn’t provide solid evidence against nested hierarchy as evidence for common ancestry. All it talks about is how mutation is not perfectly random, causing homoplasy. I don’t think anyone disputes that mutations are not perfectly random, nor that homoplasy does not occur, so Borger is arguing against a straw man.

And this doesn’t change the fact that trees produced by common ancestry have objectively different topologies than those produced by separate ancestry, nor that real-life trees look much more like the trees produced by common ancestry.

@Meerkat_SK5

“Furthermore, we show that our proposed model accounts for most of the mutations at neutral sites but it is probably the predominant mechanism at positively selected sites. This suggests that evolution does not proceed by simple random processes but is guided by physical properties of the DNA itself and functional constraint of the proteins encoded by the DNA.”

Evolution: are the monkeys’ typewriters rigged? | Royal Society Open Science (royalsocietypublishing.org)

If you think that this quote is evidence against common ancestry, then you must share in @colewd’s delusion that common ancestry is interchangeable with unguided evolution.

(Not to mention that the second quote isn’t evidence for guided evolution either – it merely states that mutation and selection are affected by codon bias and functional constraint, which I don’t think anyone has disputed.)

[John_Harshman]

In other words, Bill has no idea what you’re doing or how phylogenetic analyses work, but he’s sure that it supports creationism.

[Dan_Eastwood]

@Meerkat_SK5

Oh please, all you did was refute a strawman argument. Common design can explain and predict those nested hierarchies WITHOUT violating those patterns. You just choose to assume it was common descent instead.

This is not correct, and we have previously discussed the reasons why.

Evolution predicts a nested hierarchy, common design does not. We might falsify a prediction from evolution if we do not find a nested hierarchy, but we can never falsify design. A mischievous designer could mimic exactly what we predict from evolution, and so evolution could be the Designer.

Can you test a hypothesis that would demonstrate the Designer is anything other than evolution?

[misterme987 (Andrew P.)]

@Meerkat_SK5

Oh please, all you did was refute a strawman argument.

In what way? I was refuting the creationist claim that differing levels of similarity can produce a nested hierarchy structure like common ancestry can. I even provided quotes from all of the major creationist organizations where they say that this is what they believe, so this certainly isn’t a straw man. Since random sequences also have differing levels of similarity, the fact that they don’t show nested hierarchy refutes this claim.

@Meerkat_SK5

Common design can explain and predict those nested hierarchies WITHOUT violating those patterns.

Do you have a testable hypothesis for how common design can explain those patterns apart from common ancestry? Remember, I have nothing against design, it’s separate ancestry that I think has been disproven.

@Meerkat_SK5

You just choose to assume it was common descent instead.

No, I’m not “choosing to assume” that, that’s the whole point of this thread.

[colewd (Bill Cole)]

@misterme987

No, you’re piggybacking on (and misinterpreting) my comment on the other thread where I said that differing levels of similarity could possibly be explained by separate creation events at different times in history. The entire point here, which I proved in the OP, is that differing levels of similarity aren’t the same as nested hierarchy.

Hi Andrew
We are not talking about random patterns we are talking about the same pattern slightly modified. Why do you think common descent is forming a nested pattern? The mechanism appears to be two copies of the same sequence diverging, When this happens across several species at different times it appears to form a nested pattern. What am I missing?

[Meerkat_SK5]

@Dan_Eastwood

Evolution predicts a nested hierarchy, common design does not.

I have already explained in post 5 how Von Nueman model does predict the same patterns.

@Dan_Eastwood

Can you test a hypothesis that would demonstrate the Designer is anything other than evolution?

:
Yes, Stay tuned on this one

@misterme987

Do you have a testable hypothesis for how common design can explain those patterns apart from common ancestry?

I do. But, I need permission to showcase it here. Otherwise, I will refer you to another topic I am discussing with @John_Harshman right now.

[thoughtful (Valerie (non-scientist))]

@misterme987

It’s programmed in Javascript, so for each nucleotide in each random sequence, I use Math.random() to generate a pseudo-random number between 0 and 1. If the number is <0.25, the nucleotide is an “A,” if it is between 0.25 and 0.5, it is a “T,” etc. This creates essentially random sequences of pre-specified length.

Thanks for the explanation.

@misterme987

I understand that this isn’t exactly ‘common design,’ but there is no way to effectively model ‘common design.’

Hmm… thinking about that a bit as I read through the thread. Try a scenario like this: create many sequences that almost all the kinds have in common, some that a few each have in common, as well as private sequences. Mutate at random* (*though I suspect not all mutations are random).

It’s late, so now I’m not sure that makes sense. But it seems to me that the randomness you’ve applied to your first scenario and the random sequences of the kinds for separate ancestry are almost the opposite of a common design scenario. I think that may be what @colewd is saying too.

@misterme987

The point is that many creationists argue that differing levels of similarity produce the same nested hierarchy as common ancestry.

The same? Even common ancestry doesn’t produce the same nested hierarchy all the time with all genes or traits. Sorry, not taking the time right now to read the full context of the creationist articles, just seems on the initial read the quotes are just saying you get a nested hierarchy if you’re grouping things from one category.

@misterme987

Perhaps you have a testable hypothesis for how a separate ancestry model can explain nested hierarchy?

Not at the moment.

This is an aside, but thank you again for being nice, detailed, and clear. (Attributes that are rare online.)