EBI — listed fastqs that are not served, and fastqs derived from SRA Lite¶

Status: draft, not filed Archive: EBI ENA Checked: 2026-10-23. Both directories below are still served as directories, and SRR11442250's fastq still carries ? for every quality. Evidence: pathologies/ena-serves-a-listed-fastq-wrongly.md, pathologies/ena-fastq-is-lite-quality.md, pathologies/ena-fastq-splits-the-barcode-read.md Where to file: the ENA support form at https://www.ebi.ac.uk/ena/browser/support

Regenerate the evidence before filing:

uv run python scripts/ena_listing_census.py \
  --census data/surveys/sample_3000@reprocessed-srr/run-to-file_ena_fastq.csv.gz \
  --out /tmp/ena-listing-recheck
uv run pytest -m pathology tests/pathologies/test_ena_serves_a_listed_fastq_wrongly.py \
  tests/pathologies/test_ena_fastq_is_lite_quality.py tests/pathologies/test_ena_fastq_splits_the_barcode_read.py

Draft report¶

Subject: three fastq_ftp files are not served as listed; some derived fastqs carry SRA Lite qualities

We read ENA's run files for a seeded draw of 3,000 runs from a corpus of public 10x single-cell data. We found three separate problems with the derived fastqs. In each case the filereport listing looks complete.

1. Listed files that are not there¶

We sent a HEAD request for each of the 4,466 fastq_ftp links listed for 2,844 of those runs. Two do not serve the file filereport describes:

Run	File	`fastq_bytes`	What `ftp.sra.ebi.ac.uk` serves
SRR11040464	`SRR11040464_1.fastq.gz`	583,713,371	301 to `…_1.fastq.gz/`, an empty directory
SRR27782870	`SRR27782870_1.fastq.gz`	606,656,529	301 to `…_1.fastq.gz/`, an empty directory

curl -sI https://ftp.sra.ebi.ac.uk/vol1/fastq/SRR110/064/SRR11040464/SRR11040464_1.fastq.gz | head -1

The other file of each pair is served correctly.

We also noticed two transient behaviours, which we are not reporting as defects:

The server answered 403 for 81 links under concurrent requests, and 200 for 80 of them when asked one at a time.
For a while on 2026-10-23, it answered 4,096 bytes for ERR15320482's _2.fastq.gz. That file is listed at 7,197,985,953 bytes and has the same Last-Modified.

2. Fastqs derived from the SRA Lite object¶

For 122 of 452 NCBI-submitted runs we read, every quality in ENA's fastq is ?, and the same is true for 22 of 278 DDBJ-submitted runs. That is the signature of an SRA Lite object. NCBI's Normalized object of the same run has the sequencer's qualities. Example:

curl -s -r 0-20000 https://ftp.sra.ebi.ac.uk/vol1/fastq/SRR114/050/SRR11442250/SRR11442250_1.fastq.gz | gzip -dc 2>/dev/null | head -4

3. A spot split through the barcode read¶

SRR14522447's spots are 8 + 28 + 91 bases in NCBI's Normalized object. ENA's two fastqs split them 28 + 99, which joins the index read to the first 20 bases of the cell barcode. All the bases are there, but no read starts with the barcode. Two files are listed for a PAIRED library, so no listing check notices.

What would resolve it¶

Listed files that are not there: re-derive or remove the two files, and check fastq_bytes against what the server holds.
Lite-derived fastqs: derive the fastqs from the Normalized object where NCBI has one.
The split spot: split by the read boundaries NCBI records for the run.