Re: Part 1 (of 3): What are major aspects of evolutionary theory?

anon1@xxxxxxx wrote:

>>>>You understand that this means we can't ever reconstruct the phylogeny
>>>>of any group without a fossil record, or with one that's poor enough
>>>>that there might as well be none? And yet we do that all the time.
>>>
>>>For this purpose, I allow living bodies as "fossils" with zero million
>>>years before current as timestamp.
>>
>>How would that help you at all? All the living bodies would have the
>>same time stamp, and you would have no way of telling which one was the
>>outgroup using your method.
>
>
> That's correct. But if I have an unrooted tree, all leaf nodes
> currently living species (timestamp 0 Mya):
> Ozziephyta---+--------+------Harrietphyta
> | |
> Davidphyta +------------+----Eddiephyta
> | |
> Rickyphyta----+ Halfphyta
> |
> Riddlephyta
> If somehow I can determine that the branching node adjacent to
> Ozziephyta is somewhere around 200-230 million years old whereas the
> branching node adjacent to Harrietphyta is only around 65-70 million
> years old, then I can conclude the link between those two nodes (or
> most of it anyway) is directed to the right (toward the
> Harrietphyta-and-four-others clade), so the root must lie somewhere on
> one of the three links from that near-Ozziephyta branching node, and
> probably very close to that branching node in any case. Because
> Ozziephyta and Davidphyta are modern, whereas the branching node is so
> very very old, the idea that either Ozziephyta or Davidphyta is an
> extremely long-lived species which is ancestral to the branch point
> would have to be rejected. Likewise, the age difference between the
> near-Ozziephyta and the near-Harrietphyta branching nodes would
> eliminate anything on the right part of the tree being ancestral to
> anything on the left part.

This is too confused to interpret properly. None of these taxa are,
presumably, ancestral to any others. It doesn't matter how old any of
the terminal species are; that has nothing to do with the lengths of the
branches connecting them with the other species. Let's not confuse
species with lineages. I'm not sure what you mean by "branching node
adjacent to". Probably you mean "internal node closest to". But how
would you determine this? What point is there in all these 0 Mya time
stamps? Are you applying some kind of molecular clock? If so, that's one
way to root a tree.

> If I discover a fossil species which is indistinguishable from the
> predicted species located at a branching point, then I can conclude
> that fossil species lived at very nearly the same time as the implied
> branching-node species.

No, I'm afraid you can't. If there were a very accurate morphological
clock, perhaps you could. But there isn't.

> That's how I might pin down an approximate
> timestamp for a branching node.

There are good reasons why paleontologists don't do this. If you want to
find out why, there is no real substitute for the actual scientific
literature, i.e. physical paper, i.e. a library.

> If in addition to the fossil species
> identical to the predicted branching node, I find several fossil
> species, each very close to some point interpolated between two
> branching nodes along that link, and their timestamps are a good fit to
> a monotonic sequence of times, that's even better evidence that I have
> the timing of that branch correct, hence the timing of the branching
> nodes themselves correct.

I shouldn't be too hard on you, because this is exactly what
paleontologists often used to do in the bad old days, before we
understood what fossils can and can't tell you. So you are not obviously
ridiculous here, just at least 30 years out of date.

>>you can determine the phylogeny about groups with no fossil record.
>
> Well if somebody gives you the answer, and you trust them, then of
> course they've done the work for you and you don't actually have to
> determine it except by saying "yes sir" to agree with them..
> For example, if somebody tells you a particular group is a clade, and
> tells you some other group is not in that first group, then any
> unrooted tree including all of that clade plus the other group can be
> rooted between the clade and the other group. But how do you figure out
> what is a clade and what is not in that clade in the first place?

I have tried to explain this a bit elsewhere.

>>Nobody is talking about inferred species, so I don't know what you mean.
>
> I talk about inferred species once in a while. For example, if you have
> three observed species, then you have an unrooted tree connecting them
> all, and there's an inferred species at the branching node of that
> tree.

Bad name to use. I would suggest the more standard term "hypothetical
common ancestor" or "internal node" or "ancestral node", depending on
what you want to tsay.

> If some DNA sequence is the same in at least two of the three
> observed species, then you can presume with high confidence that same
> sequence or something very close to it is in the inferred species.
> (There might be one or two random mutations that by chance exactly
> match in position and base-change in two different branches despite not
> being present in the branching node that joins them, but it just
> wouldn't happen that a large bulk of by chance exactly matching
> mutations would occur in two different branches.)

That depends on the lengths of the branches. There are algorithms for
figuring out the probability of this, given the tree and a model of
evolution.

>>>How do you build that very first foundation, and be reasonably sure you
>>>got it right?
>
> (Repeating that question because you keep evading answering it.)
>
>
>>The first fossils of chordate non-vertebrates ... Yet long before
>>that the chordates had been recognized as a distinct group. So were
>>vertebrates, mammals, and so on, long before there were fossils that
>>might have been used to root a tree. How do you think that was done?
>
> By guessing. Some of the time the guesses are correct. Chordates really
> do include verbebrates, which include mammals. But some of the time the
> guesses are not correct. Fish and amphibians and reptiles and mammals
> and birds are not separate clades as originally believed (except
> mammals and birds which are indeed separate clades).

Actually, they were not precisely believed to be separate clades. Early
systematists didn't really care whether they were clades or not, as long
as they were in some way distinctive. But anyway, that was a good place
to start. You can determine all this without fossils, by the way. The
paraphyly of "fish" doesn't need fossils to find, just an outgroup. And
you can keep adding outgroups if you don't like the outgroups you have
already. You will eventually be forced to the conclusion that either
(for example) mammals are a clade, or there is a clade that includes
both koalas and eucalyptus, but not cats. I suppose that at bottom
that's a crude form of evolutionary clock argument, but it's the best we
can have. The fossils just don't save you, because they require other
assumptions you aren't thinking about, that are considerably more tenuous.

> I'm asking how you
> use evidence to decide which guesses are correct and which aren't, or
> equivalently how to re-build the cladogram from scratch with only
> evidence, no guesses, and then compare what you get now with what you
> guessed before to find all your guessing mistakes.

I suppose that the way we do it is to rebuild only a subset at a time,
and test that against the data. This requires that some of the previous
ideas not be false, not some of them can be.

>>And the tree is rooted by outgroup; you pick a non-animal or three as
>>outgroup. In this case, good evidence suggests that choanoflagellates
>>are the closest outgroup to animals.
>
> What evidence gives you choanoflagellates as an outgroup of animals,
> instead of a degenerate animal within the animal clade?

Could be. If you don't like that, try another protist, or whatever.
Eventually you will end up with all of life inside the animal clade, and
E. coli will be "degenerate animals" too. If you're cool with that, then
there is in fact no way to root the tree by outgroup.

> To the point: How do we know that species that form a blastula are a
> true clade, that there's not a single non-blastula species that
> degenerated from a blastula species?

If it were just a single species, the tree would show us. In order to
make animals not a true clade, you would need for *all* species that
don't form a blastula to be degenerated from a blastula species. That
is, if you have a tree on which blastulas are on one part, separated by
a branch from all non-blastulas, then either the blastulas form a clade
or the non-blastulas form a clade. At this point you have to decide
which is the more credible clade. Fossils won't help, and are
unnecessary. If you are willing to believe that prokaryotes are
degenerate animals, there is no solution. (Except for invoking molecular
clocks, self-polarizing characters like SINE insertions, or gene
duplications. Those would work, but they aren't necessary.)

> Many forms of cancer are derived
> from human species via mutation. What if a cancer broke free from a
> body and learned to live independently of its original host, and was
> recognized as a new species of protist, and we haven't gotten around to
> sequencing it because we have tens of millions of other species ahead
> of it in the queue? What if there are a hundred differnet species of
> freed cancers living as protists nowadays, and the choanoflagellate you
> mistakenly chose as an outgroup happened to be one of them?

Then you would get a wrong answer. Try another outgroup and see what you
get.

>>An unrooted tree is as good a bit of evidence for evolution as a
>>rooted tree.
>
> How exactly does an unrooted tree show evidence of evolution?

It's a nested hierarchy. Separate creation doesn't predict an unrooted
tree in exactly the same way it doesn't predict a rooted tree. It's the
robustness and objective existence of the internal branches that really
count, not their time-directions.

> Suppose there's a mountain range (the Rocky Mountain "continental
> divide") separating the Pacific drainage from the Atlantic drainage,
> with branches off from the main ridge between each group of rivers,
> like one branch-ridge between the Columbia and Sacramento river
> systems, and another branch-ridge between the Rio Grande and the
> Mississippi river systems. Now suppose some space alien sets up
> microwave relay towers at all the local maxima along those ridges. Now
> we discover those microwave relay towers, and notice they form an
> unrooted tree, and we conclude they evolved??

Local maxima? Branch ridge? If I can decipher your statement, it's an
analogy between rivers and trees. But rivers *are* rooted; they have a
clear direction of flow. So if anything, you have shown that rooted
trees are useless in showing evolution. I think you have shown nothing
except that there no point to this analogy.

>>Just the existence of a tree is evidence of a phenomenon for which
>>creationists have no explanation.
>
> The "fitness landscape" is like mountain ridges, separated by evil
> drainage which washes away any life not on a ridge. The Creator was
> wise, and placed all Created life along the ridges, not in the wash-out
> places.

You analogy doesn't make enough sense to argue with. To the extent that
I can make sense of it, it seems to be a claim that ecological niches
are organized into a nested hierarchy, and the evidence suggests otherwise.

>>We are all agreed that the Cambrian is a very different place
>>from the Jurassic, which is itself very different from today. And that
>>the biotae of one period are more similar to those of adjacent periods
>>than to those of distant periods. That, in a gross way, is evidence for
>>evolution, and rather good evidence at that.
>
> Well, if "we" doesn't include the people who reject all evidence of
> ancient time, who insist everything was sorted in the Flood. So maybe
> there's a short route to evidence for evolution in this general sense
> for my check-list survey: Ask about crystals and radioactive dating,
> and then fossils grouped by when they were deposited in sediment, just
> in a general way as you say there. If all that is agreed upon, then
> either they agree such change happened, or they need to explain where
> the logic breaks down in their opinion.

Yes, you should ask about the sorting mechanism. None of the proposed
creationist sorting mechanisms make any sense at all.

> Re single gene that gets copied then two copies co-evolve along single
> whole-cell tree of evolution:
>
>>copy 1 tree copy 2 tree
>>A--------+------*--------+---A
>> | |
>> B-+-/ \--------+--------B
>> | |
>> C-----/ \-----C
>>You have two copies of a gene (or two related genes, whichever you want
>>to call it). You sequence both of the genes and it gives you 2 unrooted
>>trees. But put both copies into a single analysis and you get a single
>>tree, which you can objectively root somewhere on the branch connecting
>>them (*). This doesn't depend on any sort of clock or midpoint rooting,
>>and I have tried to show this by making the two genes not very
>>clocklike. Now do you get it?
>
> Yeah, that's like I had only clearer to present and makes it clear
> whether co-evolution of the two copies is supported or not.

Actually, it's nothing like you had. I'm still not sure you understand
the point.

> In this
> case co-evolution seems to be supported. More likely there will be some
> unresolved inner nodes, and one half will come out slightly different
> structure from the other due to different single-gene resolution of the
> unresolved whole-genome node or vice versa. But if the two structures
> are "close enough" given the low confidence scores for the nodes that
> disagree, that's good enough. If the high-confidence branches agree
> exactly, the hypothesis is supported.
>
> Has this exact kind of analysis been done across the three domains?

I seem to recall that it has. Didn't you reference something of the sort?

[snip]

> Side question: One of the basic principles of constructing (unrooted)
> trees from DNA sequences is that if two long DNA segments are nearly
> identical, they must be homologous, either both from a common ancestor,
> or one from the other, because a tornado in a junkyard doesn't make the
> same sequence twice. But that doesn't apply to repeating patterns that
> spontaneously fill a huge segment with a short pattern over and over,
> where the chance of a pattern of length N is roughly 1/N rather than
> 1/(4**N). I presume there's a way to detect such repeats within a
> genome, even if they have subsequently suffered a small number of
> random mutations, such as ACTCCACTCGACTCCACTCCACTCCACGCCACTCCATTCCACTC
> or even small additions or deletions? Is there a program that
> eliminates all such mutated-patterns from consideration, even if one
> such repeating pattern is inserted in the middle of an older repeating
> pattern, such as ACTCCACTCGACTTAATAATAACCACTCCACTCCACGCCACTCCATTCCACTC?

Your implicit, false assumption is that we start with whole genomes and
eliminate those parts we don't like when doing phylogenetic analyses.
But what we actually do is start with parts we are interested in, and
which prior experience has suggested might be useful. We don't sequence
whole genomes, usually, just small homologous pieces.

Now, your question makes some sense in terms of aligning those pieces we
have picked with each other. But we pick pieces that do not consist of
long repeats like this, because they are so variable and likely to be
useless in determining phylogeny (other than at very close ranges --
much of genetic fingerprinting consists of such loci, called
microsatellites).

> After that program has finished, whatever is left could then be checked
> for actual duplicates of ordinary sequences that aren't repeating
> patterns. Has this been done on the human genome? If so, how many sets
> of duplicates have been found? Same question about chimp genome. Then
> are there any duplicates in one but not the other, or more generally
> any sets of duplicates where there are more copies in one genome than
> in the other?

All you have to do is ignore such troublesome regions and use other,
more well behaved ones.

>>>Looked it up just now, some kind of spontaneous multiple repeats, can't
>>>figure out what's the difference between these and "tandem repeats".
>>
>>No, SINEs are not multiple repeats in that sense. They are usually
>>inserted one at a time, though they can be inserted at many different
>>places in the genome, so in that sense they are multiple. The idea is
>>that the point of insertion is random, and that deletions are very
>>unlikely to exactly match insertions.
>
> Ah, so whereas tandem repeats could be caused by relative slippage of
> the original and anti-copy during the process of DNA anti-copying, like
> a phonograph record that keeps jumping back and repeating the same
> short segment over and over and over, while somebody is trying to tape
> the record so the tape copy has one segment over and over and over, and
> copies of that tape likewise have the same tandem repeat, SINE copies
> must be some other mechanism? Any idea what causes it?

SINEs are a form of retroelement. They are transcribed into RNA, then
reverse-transcribed into DNA and inserted into the genome. They are a
degenerate form of LINE. You should look both of those up. There are
nice web sites that do explain the mechanism involved. A search should
locate one.

[snip]

>>So if you find a sequence in multiple species that doesn't have a
>>SINE inserted, and more species with the SINE inserted at exactly the
>>same spot, the insertion is homologous and polarizes the tree: the
>>species with the insertion are more closely related to each other than
>>to any species without the insertion.
>
> Let me try to clear that up: If you have an unrooted tree, and there's
> one link such that every leaf to one side of that link has the insert
> while none of the leaves on the other side have it, you presume the
> side with the insert is a true clade?

Yes.

> That's the exact opposite of what
> you assume normally, that if there's a big chunk of DNA present in
> every member to one side of of a link but not at all on the other side,
> you presume a deletion occurred, so the side with the deletion is the
> true clade.

Not true. You actually have no way of knowing whether it's an insertion
or a deletion in most cases. The SINE is clearly an insertion, because
we know the mechanism by which SINEs are produced.

> So is the idea that certain patterns get commonly inserted
> over and over and over in random places, but it's very unlikely a
> random deletion will exactly delete that particular pattern, although a
> random deletion *could* delete any pattern whatsoever, so if you add up
> all the apparent changes like that, scoring one way for SINE-looking
> differences and scoring the other way for all other patterns (except
> tandem repeats which you don't count at all), the balance tells you
> which way the arrow of time went?

I have no idea. I don't know what you just said. Did I answer the
question anyway?

.

Relevant Pages

  • Re: Part 1 (of 3): What are major aspects of evolutionary theory?
    ... characters against different species or individual and performs ... tree connecting all the species/individuals. ... Do you have any summary statistic, regarding the complete unrooted tree ... branches of the tree are true clades, and that the true root lies ...
    (talk.origins)
  • Re: Atheists support evolution because evolution supports their worldview
    ... In the creation account, whole varieties ... groups of things that can become many species. ... If we expect to see X when Y exists, and we don't see X, that's considered evidence that Y doesn't exist. ... of fruit on it's one tree, and one is produced every month. ...
    (talk.origins)
  • Re: Part 1 (of 3): What are major aspects of evolutionary theory?
    ... >> evoution and common descent, then I ask them what theory, other than ... >> specific tree model) preferred over the null hypothesis of random ... >> species, such that the whole thing could be processed by hand or at ...
    (talk.origins)
  • Re: Part 1 (of 3): What are major aspects of evolutionary theory?
    ... > kind of external information from another source, such as fossil ... > Just emphasize that bone structure etc. of a currently-living species ... talking about how to root the tree, not how to find the topology. ... > Suppose there's a known branch point (internal node in unrooted tree, ...
    (talk.origins)
  • Evidence for Anthropogenic deforestation in W europe during the Neolithic
    ... are both notable and result in decline in all species. ... Main conclusions While present-day populations of P. cembra are ... occurrences of this tree since the early Holocene. ... While it may seem counterintuitive that nutrients can move uphill ...
    (sci.archaeology)