plink 1.9 unusual error: file read failure

[Kaustubh Adhikari]

Hi, I'm getting an unusual error when trying to merge two files. Even tried this with the latest July 15 build and the problem persists. The command is: plink --bfile ph3thg2 --bmerge he_data2 --allow-no-sex --make-bed --out he_data_1kg

The log says:

PLINK v1.90p 64-bit (15 Jul 2015)          https://www.cog-genomics.org/plink2
(C) 2005-2015 Shaun Purcell, Christopher Chang   GNU General Public License v3
Logging to he_data_1kg.log.
Options in effect:
  --allow-no-sex
  --bfile ph3thg2
  --bmerge he_data2
  --make-bed
  --out he_data_1kg

16039 MB RAM detected; reserving 8019 MB for main workspace.
Allocated 6014 MB successfully, after larger attempt(s) failed.
2355 people loaded from ph3thg2.fam.
3425 people to be merged from he_data2.fam.
Of these, 3425 are new, while 0 are present in the base dataset.
Warning: Multiple positions seen for variant 'rs3094315'.
Warning: Multiple positions seen for variant 'rs3934834'.
Warning: Multiple positions seen for variant 'rs9442372'.
512989 markers loaded from ph3thg2.bim.
503081 markers to be merged from he_data2.bim.
Of these, 313 are new, while 502768 are present in the base dataset.
502717 more multiple-position warnings: see log file.
Performing single-pass merge (5780 people, 513302 variants).

Error: File read failure.


Each of the two individual files, ph3thg2 and he_data2 seem fine, as --missing works fine on both, and if I want to merge one of them with itself, e.g. in plink --bfile ph3thg2 --bmerge ph3thg2 --make-bed --out he_data_1kg, that works fine too. Can you please suggest what might be the reason for this error? They are from different builds hence cm and bp columns are different, but that shouldn't matter, right?

[Christopher Chang]

Well, this is probably a bug; PLINK may be trying to read off the end of one of the files due to a bookkeeping error, or something similar.

Can you send me files to replicate this crash with?  (If not, I will send you a series of debug builds, but it may take several iterations before we figure out what's going on.)

[Christopher Chang]

This should be fixed in the July 17 development build.  The issue was that --bmerge could get confused if there were two IDs that were almost identical but had different capitalization and different phenotypes (or parental IDs) simultaneously.

On Friday, July 17, 2015 at 6:06:42 AM UTC-7, Kaustubh Adhikari wrote:

[Kaustubh Adhikari]

Fantastic, thanks for the prompt resolution.