>The reason is that typing, readind microsatellites alleles and so on
>is done by another group, I collaborate with them for linkage
>computations. So I checked the pedfile for pedigree consistency
>but not for markers typing for which my colleagues assured that
>everything was OK.
I'm surprised this has never happened to you before - I get this all the time.
The standard procedure round here is:
1) Enter results from the lab
2) Check data with DOLINK (only picks up gross errors)
3) Find problems and return to lab for resolution
4) Repeat above till gets through DOLINK checking OK
5) Run through UNKNOWN, get more errors
6) Read pedigree file into PEDRAW to locate errors (because UNKNOWN doesn't
say where they are, though Alessandro is working on it, good for him!)
7) Find problems and return to lab for resolution
8) Repeat above till gets through UNKNOWN OK
9) Run through MLINK, still get some errors
10) Read into PEDRAW again, these ones are really hard to find because they're
subtle enough to get through UNKNOWN
11) Find that problem must be in one of perhaps ten different subjects which
are mutually incompatible, but no idea which one is wrong, return data to lab
with this helpful message
12) Repeat above till gets through MLINK OK, publish results
>While thinking (deeply) about this, I noticed that many
>papers with linkage data are done the same way
>I mean one group at the bench and another one at the computer.
>I wonder if this could be one of the reasons of a lot of papers whose results
>have never been confirmed or proved later to be wrong.
I think in general it's good to have the separation because it may help to
reduce certain kinds of observer bias and it may mean that if there are
problems (e.g. non-paternity) then at least they come to light sooner rather
than later. I think that the effect of such errors would generally be to
introduce a bias against the detection of linkage. I don't think they are
really likely to produce spurious false positive results, but it is an
interesting question. Suppose that a number of genotyping errors are
made at random, but then we only detect the ones which actually produce
incompatibilities and the rest get through. What would the expected
effect be? Newton Morton has been interested in this topic - he speaks of
typing errors as artificially inflating the distances of a map and says that's
why all genetic maps are "too long" for their chromosomes.
>During this period of linkage between 1994 and 1995
>I wish you everything you want and even more.
Oh no! Two puns in two postings! (The other one about being overLODed). We
thought the English sense of humour was strange....
Best wishes to all
- Dave Curtis
