Archiver > GENEALOGY-DNA > 2003-02 > 1044546574

Subject: Re: [DNA] mutations and paper trail
Date: Thu, 6 Feb 2003 10:49:34 EST

In a message dated 02/05/03 6:05:19 PM Pacific Standard Time,

> It concerned s a 12 marker Y DNA test (the context - the phylogenetic map I
> refered to explained this)
> Thanks for your (new Version?) of the mutation calculator..very useful.
> The standard rate of .002 seems to low in my opinion, based on the actual
> rate i observed and as reported by actual mutations reported tonight on the
> list and before, but maybe we are all exceptions, and on a large population
> the rate would indeed only be .002 ?

OK, the Mutation Calculator for 12 markers gives the following results:

For 20 transmission events (where you observed one mutation), the "expected"
(i.e. the average over a large number of trials) number of mutations is 20 x
12 x .002, or 0.48. Obviously, there is no such thing as half a mutation, so
you would expect some cases to have zero mutations, some cases to have 1
mutation, and even a few cases where there are two or more mutations. The
likelihood of various outcomes from the Mutation Calculator (rounded to the
nearest whole percent) is:

0 mutations: 62%
1 mutation: 30%
2 mutations: 7%
3 mutations: 1%
4+: ever smaller numbers

That means it's not at all unusual for you to encounter 1 mutation in your
sample -- it will happen in about 30% of all samples if the "true" mutation
rate is .002.

For 40 transmission events, where you think you have two or three mutations,
the likelihood of observing various outcomes is:

0 mutations: 38%
1 mutation: 37%
2 mutations: 18%
3 mutations: 6%
4 mutations: 1%

When we talk about a mutation rate, we aren't talking about a clock ticking
down so that a locus mutates every 500 generations exactly. Mutations are
completely random events, and the cells producing sperm have no memory about
whether a mutation took place in the preceding generation or (probably) even
at a different spot on the same Y chromosome.

Maybe one way to visualize the process is to imagine you are entering the DNA
Casino, with 12 roulette wheel tables stretched down a row, one for each
marker you test. Each wheel has 499 black positions and one red position. You
spin the wheel at table DYS19, and if you land on red, you say you have one

You proceed to the next table, DYS388, and repeat the process. Your chance of
landing on the red is completely independent of what happened at the first

You keep on going down the row of tables, and when you exit at the other end,
there is a poll-taker asking you how many mutations you got. You tell him
none or one or two or whatever. Would he get all excited and think the
roulette table is rigged if you reported one mutation? No, but if you
reported five mutations, he'd be concerned.

Now imagine that 20 people do the same thing (20 transmission events). The
poll-taker's summary is likely to show that some people had zero mutations,
some one mutation, some two mutations, similar to the distribution from the
Mutation Calculator.

Now suppose the roulette wheel inside the casino actually has two red slots,
but the poll-taker at the exit doesn't know this. We've doubled the mutation
rate to .004, but mutations are still rare, so he would have to survey a very
large number of people before he could conclude that the number of red slots
is larger than he had been told.

I think our minds aren't designed to have very good intuitions about the
range of possibilities which can develop purely by chance. We're more likely
to focus on the "unusual" situations -- when you solicited reports from other
people who also saw mutations in their samples, you were getting a biased
result. And we're more programmed to look for patterns, for example
speculating that one branch of a family is "prone" to mutations while other
branches are more "stable." You must consider the whole data set, not just
the "interesting" or "unusual" situations. This is just as you said when you
mentioned "large populations."

Ann Turner
GENEALOGY-DNA List Administrator
DNA preservation kits: >

This thread: