On 15 Aug 2002 13:51:52 -0400, viro@weyl.math.psu.edu
(Alexander Viro) wrote:
>In article <20jnlugg9p4fm48s5r9boenefebe4h5ss1@4ax.com>,
>Philip Deitiker <pdeitik@att.net> wrote:
>
>>This is not true in the way which you have said
>>it. The basic problem with mtDNA dating is the calibration,
>> the C/H LCA may be in error. The basis of error in mtDNA
>> divergences over 100,000 ky is the calibration, all
>> other sources of error only fractionally decrease
>> confidence.
>
>Actually I wonder what's the confidence interval for their
>200kya estimate. From the text on the page in question it
>appears that
> a) distance between H and C alleles is 2
The defect in the gene probably shows up in the protein
structure; however you can tract synonymous mutations as part
of a molecular clock, also intronic and downstream sequence.
Here again I don't know what Paabo did, I looked for the
literature reference this morning. I am _EXTREMELY_ (add a few
!!!!!!) dubious of prepublish press releases and I have given
a number of critiques of Paabo here. I am however hopeful that
Paabo is consistent in the direction of error and if so his
results seem to be consistent with my previous expectations. I
really need to see the sequence info and do some alignments.
> b) that locus is under strong selective pressure in H
Think about it though, such a gene would have expanded WITH
the population or shortly their after (by my estimates ~125
kya) and reached an equilibrium level of .99999 or so in the
population. The 'gene' as functionally defined could have
quasi-equivilent alleles that randomly diverged after that
initial burst and quasi-fixation.
> d) it's autosomal
Definitely, and here again I have to defer to the MRCA of
others since there is no firm basis for MRCA ing of any gene
based on good variant genetics in apes or chimpanzees. This is
a definite problem. If you go to my site and see the
calibration page
http://home.att.net/~dnapaleoanth/MoleCalib.html
" Using Autosomal gene mutations to figure out relatedness of
groups DNA mutations can be compared across families, and a
relative distance put between them based on accumulated DNA
differences. This says nothing about selection or neutrality,
and does not attempt to explain how the method works; it is an
empirical, observable fact, and can be used without
explaination. Eastall and Herbert, in 1997 counted the
mutations (substitution rates) at "fourfold degenerate sites"
of "nuclear genome-coding DNA" Abstract- "Evolutionary
divergence times can be inferred from molecular distances if a
molecular clock can be assumed and if the substitution rate
can be estimated. We present new evidence from relative rate
tests that the rate of substitution at fourfold degenerate
sites of nuclear genome-coding DNA is uniform in primate and
rodent lineages. We also review recent relative rate test
results showing substitution rate uniformity in the nuclear
genome of simian primates. DNA distances between a range of
mammalian taxa shows that a molecular clock is inconsistent
with many assumed divergence times irrespective of the assumed
substitution rate. We find that the substitution rate that
'implies' the best 'compromise' fit with divergence times
across the range of taxa is 2.0-2.25 x 10(-9). This range of
substitution rates 'implies' a divergence time of humans and
chimpanzees of 4.0-3.6 million years ago. This postdates the
occurrence of Ardipithecus ramidus and the earliest occurrence
of Australopithecus afarensis, suggesting that the common
ancestor of humans and chimpanzees was bipedal and that the
trait has been lost in chimpanzees rather than gained in
humans." In a conference report in 2001, Goodman reported the
work of his group and others in studying the phylogentic
relationships of humans and other members of the primate order
using non-coding non functional nuclear DNA loci.. The
substitution rate of these sequences is relatively fast,
allowing Goodman and collaborators to draw up a phylogentic
tree up using their primate data sets (41 of 60 primate
genera, comprising 65 primate species in total.) He used
fossil evidence to 'anchor' the calibration rate. From his own
(Goodman et al 1998) and previously published work (Shoshani
et al., 1996) he determined that "The phylogenetic evidence
from these datasets representing different nuclear loci were
mainly congruent with each other as well as with extant and
fossil osteological evidence".
Accordingly he constructed a phylogeny based on Hennig (1966)
ideas, where-
1.each taxon held all species descended from a common ancestor
(a monophyletic group or clade)
2.heirachic grouping of lower ranked taxa into higher taxa
ought to follow the phylogenetic relationship
3.taxa of the same evolutionary age within a given taxon
should be 'ranked' equally
Goodman acknowledges that the fossil record is so poor
that it allows only a scattering of branch points in
primate evolution to be identified and a time depth
estimated. The solution used by Goodman and many other
workers (Goodman, 1986; Bailey et al., 1991, 1992; Porter
et al., 1997a, b, 1999; Barroso et al., 1997; Goodman et
al., 1998; Meireles et al., 1999; Page et al., 1999;
Chaves et al., 1999) was to develop models of 'local
molecular clocks'. These clocks 'adjusted' the rate of
molecular evolution (nucleotide base substituion) making
it slower or faster - according to whether workers
identified a particular non coding DNA lineage as being
slower evolving than others or not. Differences between
primate lineages in the rate of neutral nucleotide
substitution has been detected by various workers. The
rate has been identified as nearly twice as fast in
loriforms vs lemuriforms (Bonner et al., 1980, 1981; Koop
et al.,1989; Porter et al.,
3a). Among monkeys, those that have a more complex feeding
ecology (i.e. frugivory/omnivory
vs. leaf browsing) have larger brains (Allman,
1999) and live longer (Rowe, 1996). They apparently also have
a slower rate of noncoding DNA mutation (Meireles et
al., 1999; Page et al., 1999; Page and Goodman, 2001 ,
contra Estall & Herbert 1997, above ). Hominids have
been identified as having the slowest rates in
primates. "
Basically what they are saying is that autosomal mutation
rates vary (between loci) for any given species and even if
they didn't there is the possible problem of interspecific
mutation rate variation. To be quite frank at this point,
because this becomes now an important issue, I have created a
model of intragenic rate variation and then applied that to
the interspecific variation and I don't think that one can
statistically resolve interspecific from intergenic if one
concludes that the intergenic rates are subject to variation
as a result of evolution. Thus I think some of what is said in
the above quote is optimistic. The other issue is whether one
can use any coding loci to clock anything. But underlying all
of these is the curses of all molecular clocking curses, a
topic I have pointed to many times which is very infrequently
discussed in the human molecular evolution literature. That
issue is bleed through variation.
Let us imagine that we are trying to establish a local
clock, say the clock is based on human molecular variation.
If we assume that all human variation was focused by a
constriction, and assume fixation, in about 30-40% of genes
that assumption maybe right, but many other genes 'bleed
though' variation, the HLA loci 'poured' through variation.
OK so we shouldn't use human coalescence as an anchor, how
about chimpanzee, same problem. Basically what has happened
is that some genes MRCA to mtDNA fixation, some genes MRCA
to probably about the time of erectoid divergence, some
even may have variation carried all the way back to chimps.
So if, for example you assume the alleles between humans
and chimp will anchor at the MRCA, there is a chance at
some loci, especially certain variably selective loci that
assumption could be wrong (See 'Myth about Eve' Ayala,
1995. Science) Therefore where do we anchor genes for
setting up the calibration, particularly if the MRCA are
circularly defined? The answer is that you need to look at
literally hundreds of genes, thousands of genes to see
where there is nodalities in the fixation times, that will
give a rough guage of when constrictions have occurred in
the past, and then you need to find some means of
externally anchoring those constrictions. Once you've done
that you can go about estimating the substitution rate, and
if your 'anchoring' informatics (fossil information) is
accurate all the 'pools' of rate estimations (based on
nodality) should overlay each other.
What if they are not? One of the basic problems is that
bleedthrough variation when it occurs is not similar. One
loci may have 2 major branches bleed through, or 3 or 4 or 2
majors and 2 minor offshoots or whatever. Apriori this might
not seem all that important but now lets add in
recombination, less also addin that during the end of
constrictions and during expansions some alleles, even
ancient alleles are lost. What this means is that the source
of recombination can be lost, but not the recombinants. With
recombination and GOOD and THOROUGH population studies such
recombinatorial events may leave notable hints for a
studious observer; however with the generally 'ignored'
process of gene conversion, no hints of the parent template
may be left except those absolutely disguised in the
convertants. Therefore the 'apparent' rate of evolution of a
loci might be variation x geneconversion + the real rate of
point mutations. And since variation differs per gene, some
might the rate of evolution. Notice also that this rate will
differ between species more or less dependent on the
population dynamic. As far as I know, noone has ever
attempted to statistically reconcile the affect that
variation will have on the rate of recombination and
statistically modulate the substitution rate. Therefore this
I think is a big factor in my ongoing worry about how much
we should use autosomals. My own personal rehash of takahata
et al. can only confidently describe one thing, that the
number of fixed loci (recent) exceeds the MRCA of monoploid
(mt and Y) pairs which makes it consistent with the trend
established by those pairs, BUT since it predicts the
constriction is older than those pairs, it really pales at
telling how much older can be confidently assigned.
> e) these guys had observed an allele at distance 1
> from H. Said allele shows simple dominant
> inheritance and gives speech problems.
I think this is a repeat of the essence of b.
>Philip, any comments? IIRC you were quite sceptical about
>estimates based on autosomal loci and here we have not just
>autosomal, but one obviously not neutral...
What can I say, I am hopeful that something meaningful can be
made of this but I really have to see what Paabo has done. All
these discussions about Paabo's work in the past now take on
relavance. If he has followed his past pattern then I expect
it to be faulty as a result of calibration and I am probably
going to have to pull some strings to get the sequence info.
If that is the case its not too bad. But them one really needs
to pair this with other recently fixed loci to determine
whether it is evolving faster or not or whether the
substitution rate is aberrant relative to fortuitously fixed
loci. Philip [pdeitik at bcm.tmc.edu]
http://home.att.net/~DNAPaleoAnthro