Re: PGK differences from horse to human

On 27 Oct 2003 19:36:11 -0800, [EMAIL PROTECTED] (George) wrote:

>I think this is what the author meant:
>>>
>Consider a selected small piece of PGK.
>This piece has 27 amino acids, the total PGK having about 400 amino
>acids.
>In this Selection, there are two differences between the horse and
>human.
>In this same Selection, there are twelve differences between the fly
>and the horse.
>This indicates a "difference ratio" of 6. 
>>>

agree with you



>
>You're saying that whatever piece of PGK is selected, or whatever part
>of the genomes are selected, the same "difference ratio" will appear.

no; see below


>Evidently when the Total number of gene differences (25,000?) are
>considered, there is a "difference ratio" of 6. The author carefully
>selected a small piece of PGK to match this Total ratio. He certainly
>didn't pick a random piece. If you consider the first sixteen amino
>acids of the Selection, there is a ratio of 10. You can probably find
>other numbers in other selections.


see your previous point.

I suspect... he knew that 6 was "the right answer" (from analysis of
large amounts of data), so chose a small stretch that showed this
ratio, so he could show you how the analysis works for specific data.
As I have said before, his example in the intro is merely to
illustrate the method. As both of us have said over and over, it is
too small a piece to be statistically significant on its own. But it
illustrates how the method works. Actual conclusions are drawn from
using large amounts of data.


>
>I think it would have been much clearer if the author included
>reference to the Total first and show the 27 piece as an example.

I didn’t write the book, so won't defend it. There is usually more
than one way to present things, and people differ in what they prefer
or find clearer.

But one cannot really talk about the total as a whole. The analysis
must be done gene by gene.

>If there are 25,000 differences, wouldn't there be at least that many
>copy mistakes? Some mistakes would not be carried forward in the
>population (damaging, neutral, etc.).

Well... The only mistakes one analyses are the ones that are carried
thru. More or less by definition. 

But it is the neutral ones that actually are the best for a clock. If
the purpose is to measure time, you want a clock that runs smoothly.
As you note, detrimental mutations are lost; similarly favorable ones
get enriched. So either of these distort the clock. Neutral mutations
are a smoother clock.

Again, the more data one uses, the less impact of odd things -- and
the better the conclusions. But discussion of the principle is easiest
with a small simple example.



>
>The author said "Make the additional assumption that it took some 40
>million years, on an average, for one mistake to be made…"
>(Introduction)
>He evidently is referring to the two mistakes in the Selection.
>What basis is there for this average number?  

I would assume that it is the number that best fits the data. It
certainly seems to here. If the fossil record suggests 80M years, and
there are 2 changes, then it is 40M years per change, for this small
data set.



>Why can this average not be assigned to all the mistakes in Total? 

Because different genes drift at different rates, as he explains (p
103) and as we have said before.

bob