GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2007-07 > 1184036609


From: "Joe Knapp" <>
Subject: Re: [DNA] Help please with marker 458
Date: Mon, 9 Jul 2007 23:03:29 -0400
References: <008701c7b415$794b0120$2401a8c0@your447023ae6b><REME20070703220030@alum.mit.edu><a81622ac0707041809y68277e45iaa87951637aacb9a@mail.gmail.com><REME20070705181857@alum.mit.edu><a81622ac0707051724y16754956obfb853b6686e74ed@mail.gmail.com><REME20070707152459@alum.mit.edu><a81622ac0707071304x11373a0en3b0b4e52a7b2e3c@mail.gmail.com><REME20070707183721@alum.mit.edu><a81622ac0707071637o34a3421dr36660529da9f75a4@mail.gmail.com><REME20070709202718@alum.mit.edu>
In-Reply-To: <REME20070709202718@alum.mit.edu>


On 7/9/07, John Chandler <> wrote:
> I have run the haplotypes through the mill. The mutation model
> curves look just as messy as before.

Well, I thought they'd be cleaner. However, I can tell you this: the
average genetic distance in the population as a whole is greater in
the non-interbreeding case compared to the single tribe case, so there
is definitely more variability in the former.

> I think the key parameter
> here is the number of men from the original batch of 5000 who
> have descendants at the end. I'm guessing that the new run has
> slightly fewer surviving lineages, but not significantly fewer.
>
> Here's what I'd expect to produce a more realistic result:

Just as an aside, is there anything in this model that specifically
breaks your method? I thought it was not supposed to be sensitive to
population structure, or whether the population is constant or
exponentially growing.

> instead of setting the population quota to a fixed number, use
> an exponentially increasing value that starts at 2 and goes to
> some limit (but add a small constant number to the nominal quota
> so that the initial pair will have children). This might help to
> damp out the oscillations, among other things.

I see no oscillations such as you brought up. The average number of
offspring per male who has any offspring varies about +/-1 from the
mean 4.3 (taking all males born in a given year) & not in any periodic
way. Looks like statistical noise, since in a population of 10,000,
only about 125 males are born, and only about half of them have
progeny during their lives.

I'll think about what I need to do to the program to do what you
suggest re an expanding population. One issue is that I'm limited by
memory concerns to about 2,500,000 people total in the simulation, so
an exponentially increasing population will hit that limit pretty
quick.

BTW, re decescendants of the original population, as far as haplotypes
go anyway, the founding haplotype at least is still the MRCA at the
end of the run. Ten thousand years isn't long enough for a new "Adam"
to rise in this model. In another test, I ran a small population of
1,000 for 50,000 years and a new Adam came up in about year 25,000.
For a population of 10,000, the average time interval would be even
longer (less drift).

> Here are the results of the analysis. I see that the biggest value is
> 0.011 this time, instead of 0.014, but let's not forget that there are
> only 29 markers now instead of 37.
>
> Scaled to avg 0.00479
> Rate --- Std dev
> 0.00648 0.00040
> 0.00224 0.00012
> 0.00480 0.00042
> 0.00461 0.00044
> 0.00817 0.00016
> 0.01126 0.00109
> 0.00593 0.00030
> 0.00149 0.00016
> 0.00856 0.00073
> 0.00285 0.00027
> 0.00696 0.00036
> 0.00820 0.00052
> 0.00541 0.00048
> 0.00427 0.00025
> 0.00571 0.00017
> 0.00423 0.00062
> 0.00218 0.00019
> 0.00269 0.00011
> 0.00619 0.00011
> 0.00361 0.00035
> 0.00276 0.00044
> 0.00206 0.00017
> 0.00733 0.00028
> 0.00500 0.00022
> 0.00577 0.00040
> 0.00116 0.00018
> 0.00366 0.00021
> 0.00268 0.00027
> 0.00265 0.00027

Looks like same as before there is general correlation, with some outliers:

http://coolohio.com/dna/jc29.jpg

Joe

This thread: