KING relatedness inference working way too long...

Question

[German.M.Demidov]

Dear all,

I run KING tool to infer relatedness analysis between 15K samples. However it works for several days in a row and still not a single bit of the result was produced.

https://www.kingrelatedness.com/manual.shtml however in the manual it says that the analysis should take seconds.

I was able to run king before, I don't understand what is wrong. No error message is produced (this is the worst). Plink recognized 600K variants in my VCF and correct number of samples, no complaints from this side.

What can be wrong? How can I check if KING actually does something and simply needs more time?

[Lemire]

No I don't think that's normal behaviour. I just ran a test with 5800 samples and 9M SNPs and king --relate (v2.1.6) took 5 minutes. Do you know how much memory you need versus how much you have? They suggest a calculation, but my test below actually used 30% more (17G versus their 13G ballpark estimate). You should also get a % progress on screen.

  KING starts at Tue Jun  1 13:24:43 2021
Loading genotype data in PLINK binary format...
Read in PLINK fam file osc.filtered_vcf.hardcalls.fam...
  PLINK pedigrees loaded: 5807 samples
Read in PLINK bim file osc.filtered_vcf.hardcalls.bim...
  Genotype data consist of 8822251 autosome SNPs, 233256 X-chromosome SNPs
  PLINK maps loaded: 9055507 SNPs
Read in PLINK bed file osc.filtered_vcf.hardcalls.bed...
  PLINK binary genotypes loaded.
  KING format genotype data successfully converted.
Autosome genotypes stored in 137848 words for each of 5807 individuals.

Options in effect:
        --related

Sorting autosomes...
Total length of chromosomal segments usable for IBD segment analysis is 2671.7 MB
  Information of these chromosomal segments can be found in file kingallsegs.txt

Within-family kinship data saved in file king.kin
A subset of informative SNPs will be used to screen close relatives.
Relationship inference across families starts at Tue Jun  1 13:28:59 2021
1 CPU cores are used...
  Stages 1&2 (with 32768 SNPs): 294 pairs of relatives are detected (with kinship > 0.1250)
                               Screening ends at Tue Jun  1 13:29:01 2021
  Final Stage (with 8822251 SNPs): 291 pairs of relatives (up to 1st-degree) are confirmed
                               Inference ends at Tue Jun  1 13:29:02 2021

[German.M.Demidov]

Thank you a lot! Yes, I think I have to limit the amount of RAM then. It is certainly a wrong behaviour =)

It was weird - I'd expect at least a warning from KING.