gender check using sequencing data

[kjaja]

Hi All,

I have exome sequencing data from siblings and would like to confirm their gender and relationship using genetics to make sure they are in fact siblings. I also want to confirm that the gender information is correct. Are there tools out there that can do this using exome seq data?

Thanks,

[vivek_]

To verify pedigree this post might help:

http://cphg.virginia.edu/quinlan/?p=300

To check gender, may be you could check the number of aligned reads to Y chromsome?

[jgibbons1]

I would second that -- map reads against a panel of Y-chromosome genes/exons.

[xied75]

But I have pure female reads that can map a lot Y?

[LiLin]

The average sequencing depth of chrY and chrX

[balaji]

in plink there is an option to check sex using X chr, I hope you are looking for this..
plink --bfile data --impute-sex --make-bed --out newfile

[husamia]

I also use the average depth of X and Y

[bw.]

Testing sex determination with CCLE samples

I've tried using [num reads mapped to chrX] / [num reads mapped to chrY]
to determine sex in some CCLE exome-seq samples. The ratios turned out to be:

9.4 -- s1
304.6 -- s2
272.9 -- s3
168.3 -- s4
220.6 -- s5
297.8 -- s6
226.1 -- s7
257.1 -- s8
241.9 -- s9
287.0 -- s10
278.6 -- s11
260.3 -- s12
9.7 -- s13
8.7 -- s14
261.2 -- s15
279.3 -- s16
9.0 -- s17
8.5 -- s18
260.7 -- s19
297.4 -- s20
8.7 -- s21
261.8 -- s22
189.0 -- s23
147.4 -- s24
291.2 -- s25

So it looks like the difference is pretty wide -
[num reads mapped to chrX] / [num reads mapped to chrY] is < 10 for all male samples and > 100 for all female samples.

Still, I'm not sure whether these thresholds are stable across exome-seq kits, gene panels, etc. I wonder if there's a more robust way to determine sex.

[oyvindbusk]

How about using % heterozygosity on X (without the pseudoautosomal regions (X:60000-2699520 and X:154931043-155260560). In our lab, male = < 30 % and female = > 50 %.

[swbarnes2]

Gender = biological sex + culture. You don't care about people's gender, you care about their sex. (And even the biology is not black and white 100% of the time)

[bw.]

@swbarnes2 cool. never realized there was a difference.

@oyvindbusk thanks, I also tried this and ended up with similar thresholds (male < 40% and female > 50%). I didn't try to filter out pseudoautosomal regions since their coordinates differ across species and assembly versions (based on PAR coordinates at:

http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/mouse/

http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/

).


Looking at 322 CCLE samples, 233 were called Male, 73 Female, and 10 Unknown (which is >= 40% and <= 50%). Out of the 233 Male, only 5 would have been called differently with your thresholds. I will see if I can check the thresholds against a different approach. Also, a lot of the CCLE cells have copy number amplifications / deletions, so these results might be skewed by that.

Here is the distribution of nHet / nHomo for chrX in CCLE samples (I used this instead of nHet/(nHet+nHomo)). The 2 vertical blue lines are equivalent to 40% and 50% thresholds, and the 30% threshold is the red line.