FAQ¶
wget can not be found in your PATH ...¶
adaptiSeq needs wget for metadata/discovery. Install it (e.g.
conda install -c conda-forge wget) and re-run. The same message appears for any
other required tool the run actually needs.
srapath can not be found in your PATH ... on a -g run¶
A direct ENA *.fastq.gz download needs only wget. You see this only when the
run fell back to the SRA path (the file was not on ENA). Install sra-tools:
adaptiSeq checks srapath / fasterq-dump lazily — only when the SRA
fallback is actually taken — so a pure-ENA workflow stays tool-light. (A missing
srapath used to be mis-reported as "not available in all databases"; it now
gives this clear message instead.)
fasterq-dump can not be found ... during conversion¶
The run downloaded a .sra file and needs to convert it (-q, -e, or -g
falling back to SRA). Install sra-tools (as above).
A download is slow¶
- Try forcing the SRA database:
-d sra(or-d ena). - Try a different ENA transport:
-r httpsor-r ftp. - For ENA/GSA, try Aspera:
-a(needs a realascp+ key). - Raise/lower batch concurrency:
-j N, or pin it with--no-adaptive.
NCBI rate limit / slow metadata for big batches¶
NCBI E-utilities allows 3 req/s without a key, 10 with one. Export a key:
Aspera (-a) transfers nothing / authentication fails¶
- Ensure a real IBM
ascpis onPATH(a no-op stub passes startup checks but transfers nothing). - Ensure a key file exists (adaptiSeq searches the conda env and
~/.aspera). - ENA migrated DSA→RSA Aspera keys; the legacy
asperaweb_id_dsa.opensshis rejected. adaptiSeq's RSA fallback covers this automatically.
A file failed its md5 check¶
adaptiSeq retries up to three rounds, then records the Run in fail.log and
deletes the corrupt partial. Re-run the command to retry. To skip checks entirely
use -k/--skip-md5 (then re-run without -k to verify later).
How do I re-download something already in success.log?¶
Remove its line and re-run:
Does -p work with the default engine?¶
No. -p (axel connection count) applies only to --engine classic. The default
segmented engine uses --max-segments (per-file ranges) and -j (batch pool)
instead.
Can I use adaptiSeq as a library?¶
Yes — from adaptiseq import fetch, resolve, get_metadata. See the
Python API.
Is the output correct?¶
Resolution, metadata, integrity, logs, and merge are covered by a differential
test suite that checks adaptiSeq's output against stored reference fixtures. The
engine changes only how bytes are transferred, never which bytes are written.
One deliberate improvement over iseq: 3-file runs (orphan/barcode + _1 + _2)
that it mishandles are downloaded correctly.