using SNV/SNP file with spectre in tumour samples

[eesiribloom]

I am calling CNVs in tumour ONT long-reads and want to try using spectre

I assume it is preferred to input SNPs from a matched tumour sample rather than germline SNPs? Or would germline SNPs be more informative as these are many orders of magnitude more frequent?

Secondly, would it work to use more accurate illumina short-read SNV calls from a matched sample or SNV calls from the same long-reads used to call CNVs?

[lfpaulin]

Hi @eesiribloom

I assume it is preferred to input SNPs from a matched tumour sample rather than germline SNPs? Or would germline SNPs be more informative as these are many orders of magnitude more frequent?

In the current version, we use SNV data from the tumor in order to compute the CN associated with the ploidy using "perfect" heterozygous SNPs (AF 50±10%) so the VCF needs to be from the tumor. In a future release, we also use the SNV and coverage data from the control to improve the calling of both CNV and Loss of Heterozygosity (LoH) which will be released by the end of the year.

Secondly, would it work to use more accurate illumina short-read SNV calls from a matched sample or SNV calls from the same long-reads used to call CNVs?

Yes for LoH detection, but not for estimation of the coverage for the CN neutral.

[eesiribloom]

thanks for the reply @lfpaulin . If it's recommended that you use the copy number neutral state from the SNV data instead of the coverage would you recommend generating SNV calls from tumour long-reads to use in this case?