The phylogenetic tree and nested hierarchy

[SchrodingersFinch]

I wanted to continue the discussion that started in the thread AronRa vs phicomingatya Debate Analysis.

Meet grandpere with genes ABC, grandmere with abc and their children: oldest male aBC, oldest female Abc, younger male abc and younger female abc.
The oldest male and oldest female mate and have four children of there own: AbC, ABc, ABC and abc.
The younger male and younger female have four children: abc,abc, abc and abc

His is the family tree
ABC x abc
|_ aBC x Abc ---- AbC ABc ABC abc
|_ abc x abc ---- abc abc abc abc

Geneticly, the fourth child of the first couple (of the 2nd generation) is closer related to his cousins, than to his own siblings. It can happen, while you have a valid tree. The "geneticly closest"-tree will be different from the real family tree.

You could've been a bit more elaborate in your example. Are the letters alleles, or chromosomes, or what? Are these haploid organisms? In any case, I don't think you can apply this to the evolutionary tree.

The family tree shows that abc could just as well be the child from either couples. This hardly has anything to do with the phylogenetic tree. If these were species instead of individuals, two species couldn't suddenly come together to produce more species. That's directly against the definition of a species.

Can you demonstrate how a similar thing could happen with species? Let's say the species A, B, C and D are related as is shown in this picture:

Can you give a plausible natural explanation, without resorting to astronomically small probabilities, of how B could be genetically closer to C than to A?
Also, the species are complex sexually reproducing organisms, like birds or mammals. Therefore horizontal gene transfer cannot occur.

I claim, and so does evolution, that this simply cannot happen. There is no plausible way through which B could've evolved genetically closer to C than it is to A. Such an example would falsify evolution. So give it a try. The Nobel prize could be yours.

 

[Gnug215]

Good idea.

This was what I suggested a number of times to phi, but it's good that you took the initiative now the other thread is locked.

 

[Squawk]

To expand a fraction on the things that Schrodinger said, the reason that we can make statements based on probability for genetic comparison, i.e. that two organisms won't converge upon the same genetic structure, is due to the size of the set of all possible genomes. As we look at gene space (the set of all possible genomes), we note that there are far far more ways of being dead than of being alive.

Or put another way, the set of genomes that result in viable organisms is vanishingly small compared to the set of all possible genomes. Ergo, nested hierarchy, the result of a branching process. If genomes could converge with any likelyhood, phylogenetics would be much harder.

 

[phi tran]

The family tree I gave, was of individuals of a species, with each having three genes. The first gene can have allele A or allele a, the second gene can have allele B or allele b and the third gene can have allel C or allel c.
The example I gave are about haploid species. I was actually having a diploid species in mind, so sorry for the confusion here.

You know I can give you examples other than cities, that are harder to break down.
But let's examine your analysis first.

Can you give a plausible natural explanation, without resorting to astronomically small probabilities, of how B could be genetically closer to C than to A?

That's easy. Let Z be a common ancestor of A, B, C and D; and let S be a common ancestor of A and B; and T is a common ancestor of C and D.

Z
|_ S ---- A B
|_ T ---- C D

S and T can be very closely related.
B can be a subspecies of S while having diverged only little. Given enough time, there's no limit to how much A can have diverged from S.
Let C be a subspecies of T with little divergence.

So S is colse to T.
B is close to S.
C is close to T.
A is far from S.

In this scenario, B can easily be (and probably will be) genetically closer to C than B is to A.
Really, I don't think such example would really falsify evolution, as you claim it would.

 

[Squawk]

A fundamental misunderstanding arises here, in particular the notion of "distance" and "closeness".

From your example we can say the following must be true if the tree is to hold.
Anything in the genome of A and B must be derived from S
Anything in the genome of C and D must be derived from T

Anything in A, B, C or D must be derived from Z

We can think of excellent examples to highlight this. Lets go with chimps and humans, both of which diverged from a single line after the gorillas. Chimps and humans are, in evolutionary terms, equally diverged from gorillas. They are the same "distance" apart. However, a comparison of bases would no doubt reveal that one or the other is either more derived than the other simply due to 6 million years of evolution in differing sized populations and the different selection pressures that have been applied.
Which would be closer? I honestly don't know, the human population is vastly greater than that of the chimps which would tend to slow down evolutionary change, suggesting the chimp is probably more derived. However, the chimp has had selection pressures applied which may well have restricted its evolution, keeping it firmly in a niche. This is overly simplistic of course, but serves to illustrate the point.


The degree of derivation is not important, hence the notion of "distance" or "closeness" being erroneous in your example. The only thing that matters is heredity, and the subsequent derivation which we can then examine. That's where the info in my post earlier comes in, the idea that gene space is so large that no two paths will stumble accross each other.

 

[biology4life]

A and B will share markers that they inherited from S, but that arose after the divergence of T from the Z to S line, these markers will not be present in T and therfore not present in C and D.

These markers would show that A and B have a more recent ancestor than B and C do.

One of the useful things about ERVs is that they serve as such markers.

 

[biology4life]

I think we may have stumbled on why Phi' misunderstands phylogenetic trees. They are not done by comparing the entire genome and just comparing say % difference, they are made using the pattern of specific markers, such as ERVs or substitutions in the cytochrome C sequence.
The reason the trees produced are such staggereing evidence for common ancestry is that whatever markers we use WE GET THE SAME TREE!

 

[Gnug215]

I think we may have stumbled on why Phi' misunderstands phylogenetic trees. They are not done by comparing the entire genome and just comparing say % difference, they are made using the pattern of specific markers, such as ERVs or substitutions in the cytochrome C sequence.
The reason the trees produced are such staggereing evidence for common ancestry is that whatever markers we use WE GET THE SAME TREE!

This is exactly what Squawk and I were talking about on Skype just earlier. This is the kind of thing that needs to explained; how exactly the tree is built and what the markers are. The markers!!

 

[borrofburi]

I think we may have stumbled on why Phi' misunderstands phylogenetic trees. They are not done by comparing the entire genome and just comparing say % difference, they are made using the pattern of specific markers, such as ERVs or substitutions in the cytochrome C sequence.
The reason the trees produced are such staggereing evidence for common ancestry is that whatever markers we use WE GET THE SAME TREE!

I'm pretty sure he simply doesn't understand what "nested heirarchical tree" means, my evidence for this is his continual creation of heirarchical trees for inanimate objects and subsequent declaration that his heirarchical tree proves that the nested heirarchical phylogenetic tree is worthless.

 

[SchrodingersFinch]

In this scenario, B can easily be (and probably will be) genetically closer to C than B is to A.
Really, I don't think such example would really falsify evolution, as you claim it would.

I made a mistake in using the term "genetically closer". I shouldn't have rushed my response, which resulted in such a poor explanation. I apologize about that.

You are entirely correct, if by being genetically closer you simply mean having the most genes in common. But this is not enough to determine the relation between species, and not how the evolutionary tree is formed. Thank you for biology4life for pointing that out. So let me give a better example.

We have four species (1,2,3 and 4) with seven genes. Each letter represents one allele of a gene.
1. xAzcvCD
2. xAEcvrs
3. xkFcBrs
4. xGHcBrs

1 and 2 have the genes x,A,c,v in common,
3 and 4 have the genes x,c,B,r,s in common
From this we can infer that 1 and 2 must have a common ancestor with the genes x,A,v,c, and 3 and 4 must have a common ancestor with x,c,B,r,s.
We will mark with a ? the genes we don't know

Now we see that the ancestors have x and c in common. Therefore they must have a common ancestor with those genes. Furthemore, because the genes r and s are found in 3,4 and 2, they must have a common ancestor with those genes, which means the common ancestor of 1 and 2 also has r and s.

There we have it: a nested hierarchy.

Now if we found a new species with the genes xkHcvCD, we wouldn't have a nested hierarchy anymore. It would result in a closed loop like this:

Making another tree using the same logic would be impossible.

The city example you gave won't work the same way. If we started only with the last cities (Cardiff, Plymouth, Cambridge, Birmingham etc.) even if knew how far they are from each other, there is no way to form the tree from that. We simply wouldn't know which clade each city belongs to.

 

[biology4life]

If anyone is interested in learning about how phylogenetic trees are constructed, the following are good resources.

A fairly basic introduction.
http://www.wiley.com/college/pratt/0471393878/student/activities/phylogenetic_trees/index.html

A more technical description of some different mathematical techniques used.
http://guava.physics.uiuc.edu/~nigel/courses/598BIO/498BIOonline-essays/hw2/files/hw2_li.pdf

 

[phi tran]

The technical description in the second link of biology4life describes exactly the methods used for clustering anything. It's also known as cluster analysis.

I've used this method in a case assignment, to estimate correlation between stock companies, to find which companies could be linked together. Because estimates of pairwise correlation between individual companies are too unreliable. I've written a C++ program which used historical stock data to provide cluster based estimates of correlations between stock companies.
Goal of the assignment, was index tracking. Other student groups didn't use clustering methods and got poorer results.

In my cities example, I\ve added a time dimension, which is not standard in clustering methods. You can easily cluster the cities without time of incorporation.


As for SchrodingersFinch's example, when the xkHcvCD species is found, we do not create a tree with close loops. We add this new species as a relative or descendant of species xAzcvCD, because that would require two mutations. Linking it with the other line, would require at least two and up to five mutations, which is less likely.
This is actually a good example of a character based approach for finding for a tree with better likelihood.


Biology4life, in clustering species, we use DNA and every part of the DNA can serve as a marker. That includes ERVs and cytochrome C sequences. And we do not always get the same tree, using different approaches. Usually, we use all markers simultaneously to build one most likely tree.
A distance method is used to get a first approximation (because distance methods are very fast), and then we continue to look for possible adjustments which result in more likely paths of evolution.

 

[Schwobar]

Hey, Phi! I've noticed you haven't posted in your special relativity thread in awhile... Are you just trying to limit yourself to disproving one of the most widely accepted theories in science at a time now? If so I consider this a sign of maturity on your part. You're growing up so fast!

 

[Squawk]

lulz, I knew you were a troll/poe.

 

[SchrodingersFinch]

The new species xkHcvCD has exactly the same genes k and H as the species 3 and 4 do, in exactly the same locations. The probability that it would evolve the same genes in the same places, separately from 3 and 4, is extremely small.

Most organism have more than seven genes, and there are quite a bit more possible genes than the 26 letters used in this example. But even if we go by these, and assume that exactly two mutations will occur, the probability would be:
[(1/7)*(1/26)]^2 = 0.003%

Now think how small the probability is with actual genes and organisms.

 

[phi tran]

Uhm, it's just a made up example of yours.
It's still more likely than a species coming from two seperate branches at once. And with hundreds of thousands of species around, 0,003% is not that remarkably small. Odds of protein evolving is much and much smaller. So if you want to talk about chances of evolution, you better know what you are getting yourself into.

Schobar, there has been no post worthy of a response in the relativity thread. The problem still stands and remains unsolved. The front of the space craft cannot accelerate that fast without tearing apart the whole space ship.

 

[Squawk]

Uhm, it's just a made up example of yours.
It's still more likely than a species coming from two seperate branches at once. And with hundreds of thousands of species around, 0,003% is not that remarkably small. Odds of protein evolving is much and much smaller. So if you want to talk about chances of evolution, you better know what you are getting yourself into.

Bit of a hodgepodge of a post, that one. You first of all argued against genomes converging, then suggested the odds were not actually that small. Hmm.

Why did you take the 0.003% literally when it was clear that we are dealing with fictitious genes, are not discussing numbers of base pairs per gene and indeed haven't laid out specifics?

Chance of evolution? Feel free, I do know what I'm getting into.

 

[Squawk]

Sorry, but I had to giggle. I just tootled on over to pharyngula, and look at the first post

http://scienceblogs.com/pharyngula/2010/05/now_weve_got_some_big_numbers.php

Too perfect, great coincidence, must be divine intervention. Praize jebus

 

[Schwobar]

Schobar, there has been no post worthy of a response in the relativity thread. The problem still stands and remains unsolved. The front of the space craft cannot accelerate that fast without tearing apart the whole space ship.

The special theory of relativity states that measurements of time, length, and mass can all be different in different reference frames. In case you didn't know, that's what the relative part of relativity means. Since you clearly don't understand even the most fundamental tenets of the theory I'd love to recommend an introductory text on the subject to help you understand... Unfortunately, Dr. Suess never wrote about physics...

 

[SchrodingersFinch]

Uhm, it's just a made up example of yours.

I provided an example of how evolution can be falsified. Obviously it must be made up because evolution has not been falsified.

It's still more likely than a species coming from two seperate branches at once.

So what? I never said that. The point is that the theory of evolution wouldn't be able to explain it.

And with hundreds of thousands of species around, 0,003% is not that remarkably small.

Yes, but I used only 7 genes with 26 different alleles. As I said: "Now think how small the probability is with actual genes and organisms."

Odds of protein evolving is much and much smaller. So if you want to talk about chances of evolution, you better know what you are getting yourself into.

Also, Schwobar this thread isn't about relativity. Stay on topic.