Hi all,
1. I am working on a new dataset and test the relatedness of all individuals by using --genome flag in plink. Why in the result, all RT (relationships) are UNRELATED and the EZ (IBD sharing expected value) are Non Applicable (NA)? Even Z0/Z1 plot show the real case.
2. How exactly can we know the best threshold for the filter that we are going to use with the data?
Thank you
You can try to plot the IBS or IBD estimates to find the threshold to remove the outliers. See page 12- 14 for details. https://github.com/felixfan/IPGWAS/blob/master/IPGWAS3.6/manual/IPGWAS_Manual.pdf
Hi Yan-Hui Fan,
Thank you very much for your answer. I am reading the manual.
Concerning 1, I have no idea, because I never used Plink for quality control or other filtering. Concerning 2, using IBD scores as a metric, we usually removed duplicates (which could be homozygote twins or real duplicates, and there is no way of knowing) and 1 degrees relatives (i.e. parent-offspring and sib-sib), though this second stem is clearly open for debate.
Hi Frederico and Sophie,
Thank you very much for your answer. I was advancing on plotting the result considering Z0 and PI_HAT separately. Thank you for these confirmations. I was just guessing the relationship from the plot.
I would do both approaches:
1st- remove duplicates and 1st degree relatives (and even 2nd degree)
2nd- as suggested by Yan-Hui Fan, I would plot IBD estimates (i.e. PI_HAT values) and remove outliers
The threshold for outliers would be different for different human populations, that's why you sould plot it instead of deciding a threshold in advance. Smaller/inbred populations are likely to share higher IBD. In previous experiences with European groups PI_HAT values of 0.15-0.20 were used to remove excess of IBD (i.e. relatedness) while keeping the maximum number of individuals. I don't know about other groups...
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis