how to reduce the RAM when phasing a large cohort (~1200) of whole genome sequencing data with about 10Mb?

[xinkaitong]

This command consumed about 60G RAM. But it phased only 10Mb area. I still left about 2500Mb area to phase. how can I reduce the RAM when phasing a large cohort (~1200) of whole-genome sequencing data?

${shapeit4} --input chr18.vcf.gz \ --region 18:1-10000000 \ --thread 10 \ --sequencing \ --scaffold chr18_1-10000000-scaffold.vcf.gz \ --use-PS 0.0001 \ --log phased.log \ --output chr18_1-10000000.phased.vcf

[odelaneau]

Multiple advice:

• try to compact the VCF as much as possible by removing unwanted variants (e.g. monomorphic or singletons)
• reduce the chink size to 5MB or even lower
• reduce number of threads
• set --pbwt-modulo to 0.05

Hope this helps,

Olivier