non-human genome
gbnci opened this issue · 3 comments
Hello:
Just wondering whether I can run ASCAT on non-human sample WGS sequencing data (for instance canine)? I have managed to download the snp data information from online database, I think we can also generate GC content information (although have not tried yet). I have match tumor and normal pair. Not sure how to handle sex chromosomes though (or just use autosomes?). A lot of unknown for me. Any suggestions would be really helpful.
Thanks.
Hi @gbnci,
This would definitely be possible but requires some tweaks with our current ASCAT version. For now, HTS data needs to be genome-based: either hg19 or hg38. This is because:
- The logR processing (T/N ratio) puts both autosomes and nonPAR at 0, although autosomes are in 2 copies whereas nonPAR is only 1 copy in males. Therefore, it needs to be set to -1 (because gamma=1 for HTS data) here (lines 147-157).
- Since ASCAT leverages heterozygous SNPs, nonPAR won't have any. We need to artificially rescue some homozygous SNPs so BAF gets segmented at 0/1 bands here (lines 52-71).
Because we don't have a methodology in place (yet, there should be something in a few months), here is a workaround solution:
- Use the
ascat.prepareHTS
function to derive logR and BAF from HTS, but setgender='XX'
for all samples, including males. Once logR and BAF files have been generated, read the logR file and, only for males, set logR=logR-1 for SNPs in your nonPAR region (depending on your reference) and save the updated logR so the file can be fed inascat.loadData
. - Use the
ascat.loadData
but don't provide any value forgenomeVersion
(it'll beNULL
by default). Once theascat.bc
object is created and before any other ASCAT command, useascat.bc$X_nonPAR=c(VAL1,VAL2)
withVAL1
andVAL2
being the start and end position of the nonPAR region on X (if it's not on X, it would require a more complex fix).
Otherwise, a simpler but dirtier (hence not recommended) solution is to just ignore chromosome X (chrom_names=1:22
in ascat.prepareHTS
), this would only work if nonPAR is located on X though.
We'll try to implement something in the future so nonPAR can be customised but can't make any promise on a deadline I'm afraid. Stay tuned!
Cheers,
Tom.
Closing this issue for now.