stoi: no conversion error

[bw2]

I got this error on one out of ~300 samples. When I ran

REViewer --reads CDS-ulp4x8.expansion_hunter4_realigned.sorted.bam --vcf CDS-ulp4x8.expansion_hunter4.vcf --reference hg38.fa --catalog ./FXN_variant_catalog_with_0.01_threshold_off_targets.json --locus  FXN-chr9-69037286-69037304-GAA --output-prefix CDS-ulp4x8_FXN-chr9-69037286-69037304-GAA_ExpansionHunter4

the output was

[2021-01-10 18:14:17.325] [info] Loading specification of locus FXN-chr9-69037286-69037304-GAA
[2021-01-10 18:14:17.389] [info] Extracted 79 frags
[2021-01-10 18:14:17.389] [info] Calculating fragment length
[2021-01-10 18:14:17.389] [info] Fragment length is estimated to be 171
[2021-01-10 18:14:17.389] [info] Extracting genotype paths
[2021-01-10 18:14:17.393] [error] stoi: no conversion

for these input files:
files.zip

[egor-dolzhenko]

Thank you for catching this! The error was caused by a missing genotype, which prevents REViewer from determining haplotype sequences. I just updated to program to generate the following error message:

[error] Cannot create a plot because the genotype of FXN-chr9-69037286-69037304-GAA is missing

Do you think this is sufficient to address the issue?

[bw2]

Thanks, yes.
I know it's standard to generate missing genotypes in SNP calling pipelines, when there's no coverage etc. and it's not considered an error.
IIUC, EH also doesn't treat this as an error, or even a warning. The EHv4 log output for the above example was

2021-01-10T21:44:06,[Starting ExpansionHunter v4.0.2]
2021-01-10T21:44:06,[Analyzing sample CDS-ulp4x8.wgs_ccle]
2021-01-10T21:44:06,[Initializing reference /gcsfuse_mounts/gcp-public-data--broad-references/hg38/v0/Homo_sapiens_assembly38.fasta]
2021-01-10T21:44:06,[Loading variant catalog from disk FXN_variant_catalog_with_0.01_threshold_off_targets.json]
2021-01-10T21:44:06,[Running sample analysis in seeking mode]
2021-01-10T21:44:06,[Analyzing FXN-chr9-69037286-69037304-GAA]
2021-01-10T21:44:09,[Analyzing FXN-chr9-69037286-69037304-GAA-official]
2021-01-10T21:44:09,[Analyzing FXN-chr9-69037286-69037304-GAA-with-0.01-threshold-off-targets]
2021-01-10T21:44:41,[Writing output to disk]

One thought is - both EHv4 and REViewer should treat this as normal, or both should treat it as an error (or warning?). It would make sense to me if REViewer just showed an empy plot for no-coverage situations. A warning by EH and/or REViewer might also be helpful so pipelines that expect to get all genotypes could catch it and treat it as an error.

[egor-dolzhenko]

Thanks for the suggestion! Generating an empty plot (perhaps with a warning message) sounds good. I'll implement this.