Relatedness vs relatedness2 from vcftools give different results

Question

[Sharon]

Hello All How to you think about this relatedness results?

When I use relatedness2 in vcftools I got this:

vcftools --gzvcf p123.vcf.gz --relatedness2 

INDV1   INDV2   N_AaAa  N_AAaa  N1_Aa   N2_Aa   RELATEDNESS_PHI
p1      p1      47388   0       47388   47388   0.5
p1      p2      28084   0       47388   39778   0.32219
p1      p3      27242   0       47388   47505   0.287081
p2      p1      28084   0       39778   47388   0.32219
p2      p2      39778   0       39778   39778   0.5
p2      p3      28753   0       39778   47505   0.329423
p3      p1      27242   0       47505   47388   0.287081
p3      p2      28753   0       47505   39778   0.329423
p3      p3      47505   0       47505   47505   0.5

But when I use that relatedness in vcftools I got this:

vcftools --gzvcf p123.vcf.gz --relatedness

INDV1   INDV2   RELATEDNESS_AJK
p1      p1      0.428714
p1      p2      -0.187837
p1      p3      -0.343416
p2      p2      0.552425
p2      p3      -0.171243
p3      p3      0.423852

P2 is supposed to be the son of p3(father) and p1(mother) or this is what We wanna prove. It was not normal birth, a kind of artificial insemination

I am surprised that the PHI score between P1 and P1 , P2 AND P2 , and P3 and P3 is only 0.5 using both commands?

I understand that both command rely on different approaches. But not sure how to conclude whether or not to confirm relatedness. How do you think? If I trust the relatedness2, is 0.3 a confirmation for parent relationship?

Thanks thanks

[mirkocelii]

If the RELATEDNESS_PHI values
p2 vs p1 = 0.32219 p2 vs p3 = 0.329423 confirm the parent-offspring relationship, what about p3 vs p1 = 0.287081 are mum and dad siblings?

Answer 1

[ognjen011]

Hi!

The --relatedness method based on doi:10.1038/ng.608 is only appropriate for large cohorts. Although this is not properly indicated in the documentation, it is revealed in questions on SourceForge. Looking at the paper equation (6) in supplementary, the values of AF near 0 or 1 make the equation numerically unstable. I would just ignore this result.

The --relatedness2 method is based on the KING inference, and you have a very good tutorial here. You can interpret the relatedness_phi as the probability to find identical alleles when randomly sampling one allele from each heterozygous individual. So for one individual AB, and the parent AC, there is p=0.25 to choose A from both individuals. That probability is 0.5 when AB is compared to AB.

From the tutorial:

an estimated kinship coefficient range >0.354, [0.177, 0.354], [0.0884, 0.177] and [0.0442, 0.0884] corresponds to duplicate/MZ twin, 1st-degree, 2nd-degree, and 3rd-degree relationships respectively

Your result confirms the parent-offspring relationship, but could also mean a sibling.

[Sharon]

Great, thank you :)

[amycberger]

Hello, I have a similar situation to the original poster and get similar values for the relatedness_phi for the pairs in my family. In the example above, p1 and p3 have a phi of 0.287, indicating they are first degree relatives, but they are presumably unrelated. I get a similar phi when I run relatedness2 on two members of unrelated families. Any suggestions on how to fix this problem?

[kanika.151]

Try running tools on only 'PASS' variants maybe then you would see a difference.