dirk
New Member
Posts: 3 
|
Post by dirk on Jan 9, 2014 15:40:51 GMT
Hello,
I calculated genomic inbreeding coefficients for dairy cattle. For some animals I get estimates above 1 or below -1. How can I explain this?
Maybe some user have already had the same problem.
Best regards
Dirk
|
|
dirk
New Member
Posts: 3
|
Post by dirk on Jan 10, 2014 13:37:19 GMT
Hello Jian,
thanks for your post. I did not estimate breeding values. In the sense of breeding value estimation I understand negative values, because breeding values have a unit, e.g kg milk.
I used the --ibc options. As I understand it, all genotyped animals are used as base. Therefore the mean of F1, F2 and F3 is 0. Most of the inbreeding coefficients fall in the interval -1 to 1. If we have, e.g. a value -0.5 or 0.5 I would say the animal is 50% less or 50% more heterozygot compared to the mean.
However, out of 900 animals two animals have the following inbreeding coefficients:
Animal nSNP F1 F2 F3
1 43771 1.73987 -1.98467 -0.099270
2 44546 2.04208 -2.26175 -0.085948
What does these F1 and F2 coefficient tell me?
Best regards
Dirk
|
|
|
|
Post by Jian Yang on Jan 12, 2014 6:06:23 GMT
Sorry, I was reading too fast. These are the three estimators of inbreeding coefficient. Please see page 4 of Yang et al. 2011 AJHG (GCTA paper) for more details. We regard the current population as the base population so that the mean inbreeding coefficient of the population is zero and the inbreeding coefficients of individuals which are less inbred than the population mean are negative.
|
|
dirk
New Member
Posts: 3
|
Post by dirk on Jan 12, 2014 12:10:00 GMT
Hello Jian,
thanks for your answer. I understand it also in a way, that animals with negative Coefficients are less inbred compared to the mean and that animals with positive coefficients are more inbred, compared to the mean.
Classical pedigree based inbreeding coefficients are probabilities so the fall in the interval between 0 and 1.
Powell et al. (2010 (Nature Reviews Genetics 11, 800-805 )) suggest that genomic inbreeding coefficients could be interpreted as correlations. So if the are correlations, the values should be in the interval -1 to 1. In fact, most of our coefficients are in that interval. However, in all our data sets (3 data sets genotyped with 50K and one data set genotyped with HD) we have some animals or some chromosomes with estimates below -1 or above 1.
Best regards
Dirk
|
|
