Review on #2051 Add test data for custom/clustering, clustermetrics and clustervisualization Reviewed and changes requested by pinin4fjords on dbaku42's pull request Thanks for opening this @Donald Baku. AI-assisted review (Claude, on behalf of @pinin4fjords). A few things to address before this can land: Scope and description The PR description only mentions

data/genomics/homo_sapiens/popgen/clustering/

, but the diff also adds

1000g_phase3_plink2_pca/

and

1000g_phase3_small/

as fresh top-level subdirectories under

homo_sapiens/

. Please update the description to enumerate all three locations and link each to the specific module(s) the test data supports. Path organisation

homo_sapiens/popgen/

already hosts PLINK content (

plink_simulated.*

,

1000GP.chr*

etc.). Adding

1000g_phase3_plink2_pca/

as a sibling of

popgen/

rather than nesting it inside puts PLINK test data in two parallel locations. Suggest moving it under

popgen/

(e.g.

popgen/1000g_phase3/

or

popgen/plink2_pca/

) so the sibling sub-tree style with

popgen/clustering/

stays consistent. Path reference in description

params.modules_testdata_base_path + '/data/genomics/...'

would resolve to a double

/data/data/

. The base path already ends in

data/

. The modules PR should use: file(params.modules_testdata_base_path + 'genomics/homo_sapiens/popgen/clustering/test.eigenvec', checkIfExists: true) Documentation • Add a paragraph + file line to the main

README.md

of the

modules

branch describing the new content. This is the consistent ask in recent merged PRs (#2031, #2044). • Add provenance docs for the new files. The branch README requires "a short description about how this file was generated [...] either in this description or in the respective subfolder." The existing

popgen/README.md

documents data generation via `plink --simulate ...`; the new files need equivalent (e.g. for the VCF, where it was subsetted from + the bcftools/plink command; for the eigenvec / clusters / features, either generate them from the existing simulated PLINK data with

plink --pca

for reproducibility, or document that they're hand-rolled minimal examples). Optional The

popgen/clustering/

test files are hand-rolled floats. Technically valid (PR #2049: "only technical validity matters, not biological interpretation"), but a few-row eigenvec generated by

plink --pca

on the existing

popgen/plink_simulated.*

data would be reproducible from documented commands and slot in naturally with the existing simulated PLINK content. Not blocking; just more in keeping with the directory's conventions. nf-core/test-datasets

Review on #2051 Add test data for custom/clustering, clustermetrics and clustervisualization Reviewed and changes requested by pinin4fjords on dbaku42's pull request Thanks for taking the

plink_simulated

-derived suggestion - that was our optional suggestion and going for it is the right instinct, but the new files have format issues that break the consuming modules. The hand-rolled files we approved in the previous round were format-correct; the regeneration introduced regressions worth fixing before this lands. Three problems: 1.

test.eigenvec

is now in pandas

to_csv()

default format (CSV with an extra row-index column and a numeric column header), not PLINK eigenvec format. The

custom/clustering

module parses with

pd.read_csv(sep=r"\\s+")

expecting the

#FID IID PC1 ...

whitespace-separated header that PLINK actually emits. 2.

test_clusters.csv

and

test_features.tsv

use

IID

as the sample-id column. The

custom/clustermetrics

and

custom/clustervisualization

modules require

sample_id

(this is the simplification we made in the modules PR — they no longer defensively re-parse PLINK headers, just check for

sample_id

). Both modules would error with

ValueError: features/clusters file must have a 'sample_id' column. Found: ['IID', ...]

. 3.

README.md

is out of date - still says "first 6 samples" and "hand-rolled minimal examples", neither of which is true any more (files are now 200-row PLINK-derived). Inline comments on each. nf-core/test-datasets

Review on #2051 Add test data for custom/clustering, clustermetrics and clustervisualization Review approved by pinin4fjords on dbaku42's pull request Round-2 blockers all addressed — eigenvec is in real PLINK format, both companion files use

sample_id

, and the subdir README now matches the regenerated data. Approving. One small follow-up: the main

README.md

entry still describes the files as "hand-rolled minimal examples", which is stale now that they're derived from

popgen/plink_simulated

. Inline suggestion below — feel free to take it or rephrase as you like. nf-core/test-datasets

#2051 Add test data for custom/clustering, clustermetrics and clustervisualization Pull request merged by dbaku42 nf-core/test-datasets