prefetch issue with sra-tools 3.0.8 causing md5sum issues

[sgsutcliffe]

Description of the bug

Bug Description

Certain SRAs cause sra-toolkit prefetch to compare the md5sum of the downloaded .sra file to a .sraalign file which gives a false negative for the match, causing the pipeline to fail.

To recreate the bug

You can use the samplesheet:

sample,insdc_accession
SAMPLE1,SRR37205362
SAMPLE2,SRR37205362

You can see the issue by:

The API request was pulling a different SRA type of file and comparing the md5sum:
$ curl 'https://locate.ncbi.nlm.nih.gov/sdl/2/retrieve?filetype=run&acc=SRR37205362' | python3 -m json.tool

{
"version": "2",
"result": [
{
"bundle": "SRR37205362",
"status": 200,
"msg": "ok",
"files": [
{
"object": "srapub|SRR37205362.sraalign",
"accession": "SRR37205362",
"type": "sra",
"name": "SRR37205362.sraalign",
"size": 146773221,
"md5": "6cadf50190a2d50be9030ade298ad4fa",
"modificationDate": "2026-02-11T17:08:50Z",
"noqual": true,
"locations": [
{
"service": "sra-ncbi",
"region": "public",
"link": "https://sra-download.be-md.ncbi.nlm.nih.gov/vast/sra01/SRZ/037205/SRR37205362/SRR37205362.realign"
}
]
}
]
}
]
}
$ md5sum SRR37205362/*
ffc0b8bbd20fde977c4a44c81771d3be SRR37205362/SRR37205362.sra

As the output shows the SRA download was fine otherwise.

The issue is caused during modules/nf-core/sratools/prefetch/templates/retry_with_backoff.sh
Specifically the block:

# check file integrity using vdb-validate or (when archive contains no checksums) md5sum
vdb-validate !{id} > vdb-validate_result.txt 2>&1 || exit 1
if grep -q "checksums missing" vdb-validate_result.txt; then
    VALID_MD5SUMS=$(curl --silent --fail --location --retry 3 --retry-delay 60 'https://locate.ncbi.nlm.nih.gov/sdl/2/retrieve?filetype=run&acc=!{id}')
    LOCAL_MD5SUMS=$(md5sum !{id}/* | cut -f1 -d' ')
    if ! grep -q -F -f <(echo "$LOCAL_MD5SUMS") <(echo "$VALID_MD5SUMS"); then
        echo "MD5 sum check failed" 1>&2
        exit 1
    fi
fi

Potential Fix

I believe that simply upgrading the sra-toolkit containter will fix it as
sra-tools 3.0.8 reported checksums missing for columns that 3.2.1 validates fine — triggering the MD5 fallback unnecessarily

Command used and terminal output

Relevant files

No response

System information

No response