Reproducible Computation and Analysis

1. Recommended directory structure

project/
  inputs/          # source structures and raw tables
  conformers/      # generated ensembles and energies
  quantum/         # input, log, chk/fchk, gbw
  fields/          # cube, grid metadata, fragment definitions
  features/        # descriptor tables and schemas
  models/          # splits, pipelines, parameters, metrics
  figures/         # scripts and exported figures
  environment/     # versions and installation records

2. Naming and metadata

Combine sample ID, conformer ID, charge, multiplicity, method, basis, solvent, and calculation stage mechanically and keep display names separate. Store units, missing-value rules, atom-index convention, and generating-script commit in CSV or JSON metadata.

3. Electronic-structure calculations

• Software, revision, and compute resources
• Geometry, charge, and multiplicity
• Method, basis, ECP, dispersion, and grid
• SCF, optimization, and frequency convergence
• Temperature, pressure, standard state, and thermal corrections
• Checkpoint mapping across restarts

4. Cube and real-space analysis

Store the field identity, origin, three axis vectors, shape, units, spacing, fragments, isovalue, and color range. Verify grid equality programmatically before differences or pixelwise models.

5. Machine learning

Preserve raw data, fixed test set, group IDs, preprocessing pipeline, random seeds, hyperparameter search space, and metrics. Fit scaling, PCA, and feature selection only within each training fold.

6. Pre-publication checklist

1. Run generation scripts in a fresh environment.
2. Trace key tables and figures back to raw data.
3. Record manual changes in a README.
4. Check licenses, citations, and publication permissions.
5. Remove confidential structures and personal data.
6. Update software targets and review dates in Methods.

7. Regenerating and maintaining this site

Page content, bilingual pages, calculation-example links, search indexes, and structured data are regenerated by scripts. Edit the corresponding source under scripts/i18n/ or the enhancement scripts instead of changing generated output alone.

npm run i18n:generate
npm run i18n:markup
npm run verify:site

• i18n:generate: regenerate bilingual pages, examples, cross-links, search indexes, and JSON-LD.
• i18n:check: check page pairs, untranslated text, and chapter/example parity.
• site:validate: check local links, duplicate IDs, canonical URLs, JSON-LD, search indexes, and calculation metadata.