Memory size issues in snp_cor()

[kkleinoros]

I am trying to run ldpred2 with sample size of 7304 and 1422515 snps.
I am calling snp_cor with nthreads=12.
Although I am running snp_cor function on one chromosome at a time, I am still running out of memory. Any solutions?

[privefl]

So, the issue is with snp_cor(), not LDpred2, right?

How many variants on chromosome 1? Did you try chromosome 22? Which positions are you using? And what window size?

[kkleinoros]

correct. 115717 variants on chromosome 1. window size = 3 / 1000 as in the tutorial. chr22 does work.

[privefl]

How much time for chr22? How much memory for chr1?

[privefl]

Please verify the vector of positions you're giving to that function.

[kkleinoros]

chr 22 20621 snps system.time(corr0 <- snp_cor(G, ind.col = ind.chr2, size = 3 / 1000,infos.pos = POS2[ind.chr2], ncores = 12)) user system elapsed 2819.791 3.573 260.073

[privefl]

Did you check POS2?

[kkleinoros]

POS2 is 0 for all elements.

[kkleinoros]

I went back to your tutorial I see I missed

To convert physical positions (in bp) to genetic positions (in cM), use

POS2 <- snp_asGeneticPos(CHR, POS, dir = "tmp-data", ncores = NCORES)

[privefl]

Yeah, this explains it.

So, this is working now?

[kkleinoros]

yes thank you