I would like to move from microarray based genotyping to NGS for copy-number and loss-of-heterozygosity analysis. The copy-number bit should be relatively simple and I hope we’ll have done this by the end of the year, however the LOH analysis is more complex as it requires high quality genotype calls and to get these from sequencing you need 10x coverage. Or do you?
I have been mulling over the idea that we might impute zygosity from 1-3x coverage. Is it possible to take our one high quality SNP call from the NGS data and to look around it in the genome to see what other SNPs have been called in that locus? Using high-resolution haplotype maps might we be able to impute over haplotype blocks and look for LOH that way?
The idea seems relatively simple to me but I don’t really have much of an understanding of the complexities involved in the analysis or how it might need to account for population bias.
Can we get LOH and CNV for £50 per sample: If we can use 3x genome coverage then we should be able to get the diploid Human genome covered in half a paired-end 100bp HiSeq lane. If this is relaxed to 1x coverage then up to five genomes could be analysed in a single PE100 lane. At current reagent costs this is only about £200 per sample, which competes well with arrays. I have seen slides where side-by-side comparisons of copy-number on arrays to NGS are made and the results for the sequencing looked great, even at as low as 0.25x coverage. At this level of sequencing the total cost could be as low as £50 per sample!