<#1949 Add full test dataset> Pull request opened ...
# test-data
g
#1949 Add full test dataset Pull request opened by sarata00 Test data for nf-core/moleculardynamics Adds test data for the
moleculardynamics
branch, used by the pipeline's
test
and
test_full
profiles. Files • `testdata/experimental_structures/1AKI.pdb`: hen egg-white lysozyme (PDB 1AKI, 129 residues, single chain). Unmodified from RCSB. • `testdata/mdps/{em,nvt,npt,md}.mdp`: GROMACS parameter files for the full test (EM, 20 ps NVT, 20 ps NPT, 50 ps production). • `testdata/mdps/{em,nvt,npt,md}_test.mdp`: the same physics with shorter runs for CI (EM, 1 ps NVT, 1 ps NPT, 2 ps production). • `testdata/samplesheet/v1.0/samplesheet_test.csv`: minimal samplesheet (
-profile test
). • `testdata/samplesheet/v1.0/samplesheet_test_full.csv`: full-size samplesheet (
-profile test_full
). • `README.md`: describes the dataset. Testing The whole pipeline completes with
samplesheet_test.csv
in about 1.5 min on 4 CPUs using GROMACS 2022 (`quay.io/biocontainers/gromacs:2022`). Related • Pipeline: https://github.com/nf-core/moleculardynamics nf-core/test-datasets
Review on #1949 Add full test dataset Review comment by piplus2 on sarata00's pull request Hi @sarata00, thanks a lot for putting this together, and welcome to nf-core/test-datasets! 🎉 This is a really nice first contribution. It follows the "as small as possible, as large as necessary" principle well: the whole branch is about 170 KB. The structure is an unmodified, CC0-licensed RCSB entry, and the README is one of the clearer dataset descriptions I've seen on this repo. Splitting
*_test.mdp
and
*.mdp
cleanly between the
test
and
test_full
profiles is great too. I checked a few things locally. They all look good: •
1AKI.pdb
has 1001 protein atoms and 78 waters, with no `REMARK 465`/`470` records, so the "no missing residues or atoms" claim in the README holds ✅ • The step counts and output intervals in the README table match the MDP files ✅ • The samplesheet URLs point at the right paths on the
moleculardynamics
branch ✅ I only have a few small suggestions, mostly about keeping the docs and comments consistent. None of them should block merging. README 1. The EM files don't match the "same physics" description. The README says both MDP sets use CHARMM non-bonded settings (1.2 nm cut-offs, force-switched vdW).
em.mdp
and
em_test.mdp
, however, use
rcoulomb = rvdw = 1.0
, with no
vdw-modifier
and
nstlist = 1
. Using slightly different settings for minimisation is perfectly fine. Could you either update the README to say the CHARMM settings apply to NVT/NPT/MD, or align the EM files with the other steps? MDP files 1. Leftover tutorial titles and comments. The full-size files still have
title = OPLS Lysozyme ...
, but the samplesheets use
charmm27
. They also have
; Fast test: 20 ps
/
; Fast test: max minimization steps
comments, which is confusing because these are the full test files. Something like
title = Lysozyme NVT equilibration (full test)
and
; full test: 20 ps
would make them easier to tell apart from the
_test
versions. 2.
ns_type
is obsolete.
It was removed in GROMACS 2020, so
grompp
will log a note that it ignores it. It's harmless, but you could drop it from all eight files. 3. Small nitpicks (feel free to ignore): •
em.mdp
and
em_test.mdp
have no trailing newline. • Several MDP lines have trailing whitespace. • In
md*.mdp
, the thermostat comment says "modified Berendsen thermostat", but the NVT/NPT files say "stochastic Bussi thermostat". Both describe V-rescale, so it might be nice to use the same wording everywhere.
test_full
sizing
1. A question rather than a request: nf-core's
test_full
profile is meant to be a realistic, full-size run on AWS for each release. 50 ps of lysozyme is still very small for MD. That may be completely fine given the pipeline's scope! It might be worth confirming with the maintainers or in
#moleculardynamics
whether a longer production run (for example a few ns) would make the full test more representative. It would cost almost nothing in storage, since only the MDP changes. A note on the linked pipeline PR (nf-core/moleculardynamics#3) This isn't something to change here, but I noticed it while checking that the paths line up. That PR changes
pipelines_testdata_base_path
to
https://raw.githubusercontent.com/nf-core/test-datasets/
and then prepends
moleculardynamics/
in
conf/test.config
. The nf-core template convention keeps the branch in the base path (
.../test-datasets/moleculardynamics/
) and uses
params.pipelines_testdata_base_path + 'testdata/samplesheet/v1.0/samplesheet_test.csv'
in the configs. Keeping the template default will save some lint and template-sync headaches later. Both approaches resolve to the same URL, so this PR works either way 👍 Thanks again, this is in great shape! Once the README/EM wording is sorted out (or you tell me it's intentional), I'm happy to approve. 🚀 nf-core/test-datasets
Review on #1949 Add full test dataset Review approved by piplus2 on sarata00's pull request Thanks @sarata00, all the changes look great. I checked the files at the latest commit and everything from my review is addressed 🙌 (I was also wrong about the tutorial using OPLS settings: the current version uses CHARMM settings, so the README is accurate.) On `test_full`: 2.6 h on 8 CPUs is completely fine.
test_full
only runs on AWS at release time, and multi-hour full tests are normal across nf-core. 1 ns with frames every 10 ps (matching the tutorial) seems like a good, more representative choice to me. If you go with it, please also update the
md.mdp
comments and the README table. The final call is up to the pipeline maintainers, though, so feel free to check in
#moleculardynamics
too. Approving now, since this doesn't need to block merging. The
md.mdp
update can come here or in a follow-up. 🚀 nf-core/test-datasets
Review on #1949 Add full test dataset Review approved by mashehu on sarata00's pull request nf-core/test-datasets