FULL_ALN_FILE_*': No such file or directory

[nhhaidee]

Hi @zhengzhenxian,

I have just used latest version (1.0.1) of Clair3, I got error in one of my sample. Please note that there is no error with v0.1-r10 which I have been using for a long time. Can you help take a look on this advice me if this is bug/error of new release

ls: cannot access 'ERR6359501.Segment_3_PA.CY146886.clair3/tmp/full_alignment_output/candidate_bed/FULL_ALN_FILE_*': No such file or directory

Please see below log for more details and let me know if need more information

Thanks,
Hai

[INFO] 2/7 Select heterozygous SNP variants for Whatshap phasing and haplotagging
[WARNING] Cannot find any 0/1 variant in pileup output using variant quality cut-off proportion: 0.19999999999999996, total heterozygous variants: 0
[WARNING] Set low variant quality score cut-off to 0.0
[INFO] Select heterozygous pileup variants exceeding phasing quality cutoff 0
[INFO] Total heterozygous SNP positions selected: CY146886: 0

real 0m0.317s
user 0m0.238s
sys 0m0.088s

[INFO] 3/7 Phase VCF file using Whatshap
This is WhatsHap 1.7 running under Python 3.9.0
Working on 1 sample from 1 family

Resource usage

Maximum memory usage: 0.038 GB
Time spent reading BAM/CRAM: 0.0 s
Time spent parsing VCF: 0.0 s
Time spent selecting reads: 0.0 s
Time spent phasing: 0.0 s
Time spent writing VCF: 0.0 s
Time spent finding components: 0.0 s
Time spent on rest: 0.0 s
Total elapsed time: 0.0 s

real 0m0.491s
user 0m0.385s
sys 0m0.117s

[INFO] 5/7 Select candidates for full-alignment calling
[WARNING] Cannot find any low-quality 0/1 or 1/1 variant in pileup output using variant quality cut-off proportion: 0.7, total variants: 0
[WARNING] Set low variant quality score cut-off to 0.0
[INFO] Set variants quality cutoff 0.0
[INFO] Set reference calls quality cutoff 6.0
[WARNING] Cannot find any low-quality 0/0, 0/1 or 1/1 variant in pileup output in contig CY146886
[INFO] Low quality reference calls to be processed in CY146886: 0
[INFO] Low quality variants to be processed in CY146886: 0

real 0m0.337s
user 0m0.289s
sys 0m0.052s

[INFO] 6/7 Call low-quality variants using full-alignment model
ls: cannot access 'ERR6359501.Segment_3_PA.CY146886.clair3/tmp/full_alignment_output/candidate_bed/FULL_ALN_FILE_*': No such file or directory
[INFO] No Candidate found! Exit in selecting full-alignment candidates

real 0m6.411s
user 0m5.689s
sys 0m0.743s
`

[zhengzhenxian]

Hi,

Seem no candidate was found in the pileup calling. We might need more details to pinpoint the missing variants. Could you send the logs ${OUPUT_DIR}/run_clair3.log and ${OUPUT_DIR}/log to my email zxzheng@cs.hku.hk? Thanks!

Zhenxian

[nhhaidee]

@zhengzhenxian I sent you an email. Thanks

[zhengzhenxian]

Hi, @nhhaidee,

Thank you for providing the logs. The warning message indicates that the full-alignment calling was skipped due to the detection of too few candidates. This issue is similar to a previously reported issue, which can be found at this link: #108.

You might try to call a longer contig and let me know if the issue persists, thanks.

[mproberts99]

I've had the same issue recently. Makes sense that there are too few candidates. Perhaps it should still output variants found in pileup calling to the merge_output.vcf.gz? In my case, it was a small region from targeted sequencing and 2 non-reference variants were found (and 9 RefCall) during pileup, but this error caused it to skip full alignment calling and never produced a merged output vcf, therefore losing the two non-reference variants it did find

[nhhaidee]

I agree with @mproberts99 that it should still produce output merge_output.vcf.gz. It is particular important for automatic pipeline as our nf-flu pipeline, if one sample failed to produce output, the whole pipeline then stops immediately. Please note that for the same data, we have not had any problem with old version is v0.1-r10

[aquaskyline]

A fix is scheduled in v1.0.2.

[zhengzhenxian]

@nhhaidee, @mproberts99
Issue fixed in v1.0.2.

[nhhaidee]

Thanks @zhengzhenxian , we will update our automatic pipeline

[max-vdL]

Hi, I am using clair3 for a relatively small region and I frequently get no variants from the full_alignment step. This is expected and clair doesn't really crash (I still get the merge_output.vcf) but the gvcf doesn't get generated. This is not a big deal but means that I can't use the gvcf for downsteam processing as it's only present sometimes.

[aquaskyline]

@max-vdL could you share a log file with us?

[max-vdL]

Certainly! I'm aligning to an ensembl cDNA reference, so my contig only has ~5k bases.
run_clair3.log

[max-vdL]

In case anyone else has the same problem: I worked around it by lowering the --snp_min_af to 0.01 and specifying --print_ref_calls. That way both the pileup and the full_alignment approach always have something to report and nothing breaks. This leads to a much larger output of course, so it should only be used for small regions.