GRCh37-style chromosome names fail panel schema checks

[lordkev]

Description of the bug

If I use a panel csv that uses chromosome names such as 1 instead of chr1, it fails the schema check that it is a string.

eg.

panel,chr,vcf,index
1kgp3,1,/data/1KG/ALL.chr1.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz,/data/1KG/ALL.chr1.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz.tbi

results in:

The following invalid input values have been detected:

* --panel (phaseimpute-1kgp3-reference.csv): Validation of file failed:
	-> Entry 1: Error for field 'chr' (1): Chromosome must be provided as a string and cannot contain spaces

If I surround the chromosome name with single quotes it's accepted, but then produces an error that it's missing from the reference panel.

eg.

panel,chr,vcf,index
1kgp3,1,/lilnasx/data/1KG/ALL.chr1.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz,/lilnasx/data/1KG/ALL.chr1.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz.tbi

results in:

WARN: Chr : [1, 2, 3, 4, ...] are missing from reference panel
WARN: The following contigs are absent from at least one file : [1, 2, 3, 4, ...] and therefore won't be used
No regions left to process
ERROR ~ Pipeline failed. Please refer to troubleshooting docs: https://nf-co.re/docs/usage/troubleshooting

I believe this is related to this issue: nextflow-io/nf-schema#81

Since currently only CSV files are accepted as panel input, the only way I can get the pipeline to run successfully is by adding validation.lenientMode = true to my config file and modifying the panel channel creation by adding a map that manually converts the chr names to strings.

Command used and terminal output

nextflow run -r b02be7c nf-core/phaseimpute --panel phaseimpute-1kgp3-reference.csv --steps panelprep --genome GRCh37 --outdir results -profile docker

Relevant files

No response

System information

No response

[LouisLeNezet]

Hi @lordkev,
I've been confronted to this issue a little ago.
This is indeed a problem with csv file.
I would recommend to use a json or yaml file instead as they are explicit about the data type.
Unfortunately addind ' around the number doesn't really solve the problem.

You can use the following code to convert your csv to a json:

awk -F',' -v nlines=$(wc -l < $CSV) '
    BEGIN {
        print "["
    }
    { if ( NR > 1 ) {
        printf "  {\"panel\": \"%s\", \"chr\": \"%s\", \"vcf\": \"%s\", \"index\": \"%s\"}", $1, $2, $3, $4
        if (NR != nlines) {
            print ","
        } else {
            print ""
        }
    }}
    END {
        print "]"
    }
    ' $CSV > panel.json