Determining gender of donors from RNASeq data

[sdarko]

Hi all,

I have, what I think, is an interesting question/problem.

We did some single-cell RNASeq (SMARTer prep, so whole transcripts not 3' tagging) on a group of samples. In healthy controls this population of T-cells is pretty darned rare so that in order to have enough cells to sort we combined HC donors in some cases. Luckily, in the "mixed" HC samples, one donor was male and the other female. The two donors in any given mixed sample are unrelated. To further complicate matters, any "mixed" population could be anywhere from 100% donor1 to 100% donor2.

Now I've been asked to determine which single cell belonged to whom. My initial thought was that I could distinguish them based on the frequency of reads that align to the Y chromosome. Looking at the attached screenshot, that seems subpar.

I've looked into HLA identification from RNASeq reads and identifying SNPs, but neither has been particularly fruitful.

Does anyone have any suggestions?

[wdecoster]

I've done something similar, using a few genes which are gender specific, rather than only Y chromosome genes. Note that you should think about the pseudoautosomal regions.

I right now don't have the time to go in depth on this, but I hope the code in the following script can point you in the right direction: https://github.com/wdecoster/DEA.R/b...DEA/DEA.R#L198

[sdarko]

Thanks for the tip.

I grabbed the genes from your link, subsetted my expression matrix, trained a random forest classifier using the gene set and known genders (with greater then 99% out-of-bag accuracy) and then used that model to assign the unknown genders.

I think it worked well.