admixture crashes without explanation

Question

[dec986]

I have the following files that I'm trying to run admixture with:

-rw-r--r--. 1 703404669 domain users 665M Jun  2 16:42 ZZZZZTYT22.bed
-rw-r--r--. 1 703404669 domain users  32M Jun  2 16:42 ZZZZZTYT22.bim
-rw-r--r--. 2 703404669 domain users  47K Jun  2 16:42 ZZZZZTYT22.fam
-rw-r--r--. 1 703404669 domain users  843 Jun  2 16:42 ZZZZZTYT22.log
-rw-r--r--. 1 703404669 domain users  28M Jun  2 16:41 ZZZZZTYT22.map
-rw-r--r--. 1 703404669 domain users 207M Jun  2 16:37 ZZZZZTYT22.merged.vcf.gz
-rw-r--r--. 1 703404669 domain users 207M Jun  2 16:41 ZZZZZTYT22.norm.vcf.gz
-rw-r--r--. 1 703404669 domain users  11G Jun  2 16:41 ZZZZZTYT22.ped
-rw-r--r--. 2 703404669 domain users  14K Jun  2 16:32 ZZZZZTYT22.pop
-rw-r--r--. 1 703404669 domain users 135K Jun  2 16:32 ZZZZZTYT22.vcf.gz
-rw-r--r--. 1 703404669 domain users  17K Jun  2 16:32 ZZZZZTYT22.vcf.gz.tbi

and when I run the admixture admixture ZZZZZTYT22.bed 27 --supervised

I get the following error:

****                   ADMIXTURE Version 1.3.0                  ****
****                    Copyright 2008-2015                     ****
****           David Alexander, Suyash Shringarpure,            ****
****                John  Novembre, Ken Lange                   ****
****                                                            ****
****                 Please cite our paper!                     ****
****   Information at www.genetics.ucla.edu/software/admixture  ****

Random seed: 43
Point estimation method: Block relaxation algorithm
Convergence acceleration algorithm: QuasiNewton, 3 secant conditions
Point estimation will terminate when objective function delta < 0.0001
Estimation of standard errors disabled; will compute point estimates only.
Supervised analysis mode.  Examining .pop file...
Segmentation fault (core dumped)

the error message is not helpful.

and when I try to use the ped file, admixture ZZZZZTYT22.ped 27 --supervised

I get

****                   ADMIXTURE Version 1.3.0                  ****
****                    Copyright 2008-2015                     ****
****           David Alexander, Suyash Shringarpure,            ****
****                John  Novembre, Ken Lange                   ****
****                                                            ****
****                 Please cite our paper!                     ****
****   Information at www.genetics.ucla.edu/software/admixture  ****

Random seed: 43
Point estimation method: Block relaxation algorithm
Convergence acceleration algorithm: QuasiNewton, 3 secant conditions
Point estimation will terminate when objective function delta < 0.0001
Estimation of standard errors disabled; will compute point estimates only.
PLINK Input file error

the above error doesn't make sense, as I made that file with plink2

How can I get admixture working correctly?

[Ram]

You may need more RAM. Try running on a cluster node with higher memory, if you have access to one. If not, you may want to lease a high memory (say, 32GB) spot instance on AWS

[dec986]

This is a 64 GB RAM server. Curiously, this happening with only 1 chromosome (chr22), but with 1,103,547 sites for 2,504 individuals. Thanks for your help, I'll try splitting the VCF into sub-files

[Ram]

It might be a highmem server, but if you're working on a login node, it's probably shared with a bunch of people. You'd benefit from a dedicated 32G machine - check with your sysadmin.

[dec986]

Even when I shrink the chromosome file down to 20,000 lines, I still get Segmentation fault (core dumped) I don't think that this is a RAM issue. The VCF file is only 3.5 MB

[ATpoint]

Run this on a local machine first with the tiny test dataset. That will confirm it is not a cluster thing, e.g. related to working on a headnode or getting killed due to improper job submission. If the error persists then it could be due to some oddness with the VCF. Next would be to get some hands on bona fide admixture data. Do the author provide some that you can use for testing. If that all is sorted out go back to the server. By the way I stopped asking here admixture mem error, but not out of RAM because zou did not provide a lot of information on what server that is, how you submitted jobs etc. These things should be in the toplevel question.

[dec986]

I'm getting input file errors whether I use .bed or .ped on my virtual machine. Admixture appears to be a very buggy program, I really wish that there was an alternative

Answer 1

[dec986]

The problem here is that the warning message doesn't indicate very much, indeed can actually be misleading. The point is to ensure that all of the individuals in the .fam file are inside the .bed or you will get Segmentation fault. The confusing part is that the error message misleads the user into thinking that there isn't enough RAM.

[Ram]

Please accept your answer to mar this post as solved.