[#1210] intermediate inbreeding checks with check.marker

Date:
2010-11-17 10:00 Priority:
3

State:
Open

Submitted by:
Yurii Aulchenko (yurii) Assigned to:
Nobody (None)

Resolution:
None Operating System:
All

Severity:
normal Hardware:
All

Version:
v1.6-4 Component:
GenABEL

URL:

Summary:
intermediate inbreeding checks with check.marker

Detailed description
Apparently 'intermediate inbreeding' checks lead to TOO many samples dropped; need to cf PLINK and decide which threshold / test procedure to use See eg request from Ross Fraser: The issue concerns the QC process, specifically the sex check for intermediate inbreeding. For example, with our ORCADES isolate metabochip dataset, 409 of 894 people are excluded for this reason. In our other metabochip dataset, from the Edinburgh Artery Study, 15 of 733 are excluded. I've talked to Veronique Vitart at the MRC HGU about the issue, and she mentioned that previous versions of GenABEL (ie <1.4.x) do not exhibit this behaviour. I wonder if it has to do with the introduction of checks for intermediate X-chromosome inbreeding since 1.4.9? Only around 100 of 200k metabochip SNPs are on the X chromosome.

Comments:

Message
Date: 2010-12-07 21:10 Sender: Yurii Aulchenko Apparently this problem may be related to a more general one. Using 'ge03d2' data set I have noticed that library(GenABEL) data(ge03d2) sd(hom(ge03d2[,chromosome(ge03d2)=="1"])$F[male(ge03d2)==0]) sd(hom(ge03d2[,chromosome(ge03d2)=="2"])$F[male(ge03d2)==0]) sd(hom(ge03d2[,chromosome(ge03d2)=="3"])$F[male(ge03d2)==0]) sd(hom(ge03d2[,chromosome(ge03d2)=="X"])$F[male(ge03d2)==0]) sd for X is rather high. Is that something about how F is computed (e.g. male genotypes, while being hemizygous, are counted as 2 alleles). The trick may be that T_{Aj} is defined as 'twice the number of nonmissing genotypes for SNP j'. Should check that 'hom' indeed uses 'twice' for autosomes and whether fixing to 'once' for male X helps. This relates to other problem -- frequencies in e.g. summary.gwaa.data are estimated under the same assumption.
Date: 2010-12-07 08:26 Sender: Yurii Aulchenko Also from Surkakameth Mahasirimongkol: I just realize that inbreeding coef. in GenABEL, use range between 0.2-0.8. Do you think this is a stringent criteria for excluding samples. Should we use other cut off like 0.3 or 0.4?My samples have about 10% excluded by this criteria in one study.

Field	Old Value	Date	By
priority	5	2011-02-07 08:45	yurii
Severity	critical	2011-02-07 08:45	yurii