Incomplete parallel run due to "auto-debug" failed

[rpianezza]

I am trying to run parallel SINGER on five VCF files (one per chromosome). For chromosomes 1 and 5 (the two longest), SINGER does not produce any trees in the output.

I checked the log files, and it seems that in both cases only one of the 1 Mbp chunks failed, while all the others appear to run fine. The failure is due to a failed “auto-debug” for that specific chunk:

“Auto-debug failed. Contact the author for help: yun_deng@berkeley.edu”

Do you have any idea what could be causing this problem?

As additional information, I am using fully homozygous individuals in the VCF file, so there is no heterozygosity within each sample. Ideally, I would like to run SINGER on the haploid version of the VCF, but this is not supported yet. I am not sure whether this could be related to the failure, but it seems unlikely to me.

Best,
Riccardo

[YunDeng98]

@rpianezza sure, could you send me your command and program output?

[YunDeng98]

@rpianezza could you check the following things which could be the causes?
(1) Did you use the singer_master file from the .tar.gz file of version 0.1.8? If not that might be the most likely reason;
(2) Otherwise, it seems only one sub-command is failing which is the segment 30_31. Could you run this command in separate and send me the output? '''/groups/nordborg/projects/ARGs_and_TEs/tools/singer/singer -Ne 10000.0 -m 7e-09 -r 7.98e-10 -input snps_5_5_pangen.noTC.noNA -output 5/5_30_31 -start 30000000.0 -end 31000000.0 -polar 0.5 -n 10 -thin 20'''?

[rpianezza]

I found the problem. There was a region in my VCF file (just after 30Mbp) where many SNPs had ALT == "N". Removing any SNPs with N as alternative allele solved the bug. Hope this will be useful to someone else as well.

Thank you for your help!

Riccardo