MolViewSpec-based 3DEM validation visualization for IHMValidation. Replaces the ChimeraX dependency for generating structural views in the validation pipeline (issue #127).
Map and model views at depositor-recommended contour levels, fetched from the EMDB API and passed as absolute contour values to Mol*, matching the VA ChimeraX surfaceview approach (raw .map files cached locally).
| 5A1A / EMD-2984 | 5A63 / EMD-3061 |
|---|---|
 |
 |
| 9MKW / EMD-48340 | 9A9X / EMD-47736 |
|---|---|
 |
 |
These entries demonstrate the EMDB recommended-contour fetch and absolute-isovalue rendering across a range of map density scales, including the 9A9X fibril case. See CHANGES.md for the full v0.2.0 changelog and BATCH_RESULTS.md for results across all 28 benchmark entries.
Scene types
- Map-model fit: structure plus density isosurface (replaces
*_xsurface.jpeg) - Q-score colored: VA black-to-blue palette (#000000 → #0D00FF), replaces
*_xqscoresurface.jpeg - Atom inclusion colored: VA red-to-cyan palette (#7A0000 → #7AFFFF, matches VA
__floatohex), replaces*_xfitsurface.jpeg
Rendering
- Three orthogonal views (X, Y, Z) with auto-centering
- HD static images (3840x1990) via Firefox + Selenium with WebGL through Xvfb
- Interactive HTML with embedded Mol* viewer
- Density rendering from raw .map files (cached locally) using absolute contour values from the EMDB API
VA data integration
- Reads per-residue atom inclusion scores from VA JSON output
- Creates custom CIF with scores in B-factor column
- Colors structure using actual validation data, matching the
__floatohexformula used in VA
from molviewspec_validation import generate_all_views
outputs = generate_all_views("5A1A", "EMD-2984", output_dir="views/")from molviewspec_validation import (
parse_va_residue_inclusion, create_scored_cif,
cif_to_data_url, create_va_inclusion_scene,
)
scores = parse_va_residue_inclusion("emd_2984.map_residue_inclusion.json")
create_scored_cif("5a1a.cif", "5a1a_scored.cif", scores)
data_url = cif_to_data_url("5a1a_scored.cif")
builder = create_va_inclusion_scene(data_url, "EMD-2984", view="z")from molviewspec_validation import generate_validation_images
images = generate_validation_images("5A1A", "EMD-2984")
fit_plots['map_model'] = images['map_model']
fit_plots['map_model_q'] = images['map_model_q']
fit_plots['map_model_inclusion'] = images['map_model_inclusion']Run the full test suite: python tests/run_validation.py
| Test | Entries | Files | Result |
|---|---|---|---|
| HTML generation (map_model + qscore, three views each) | 28/28 | 168 | All pass |
| VA data pipeline (real inclusion scores, three views) | 28/28 | 84 | All pass |
| Edge cases (small, large, nucleic acid, low-variance, low-res) | 5/5 | 15 | All pass |
| Total | 267 | 61 tests, 0 failures |
See BATCH_RESULTS.md for the full benchmark table.
molviewspec_validation/ init.py Package exports scenes.py MolViewSpec scene builders (5 scene types) screenshot.py Headless Firefox HD PNG export integration.py Drop-in interface for em.py va_data.py VA JSON parsing and custom CIF creation emdb_api.py EMDB REST API client for contour levels tests/ run_validation.py Reproducible 61-test validation suite benchmark_dataset.csv 28 entries with metadata
- molviewspec >= 1.0
- gemmi (CIF manipulation)
- selenium >= 4.0 (static image export)
- Firefox + geckodriver (headless rendering)
- Xvfb (virtual display for WebGL support)
- Very large CIF files (>8 MB) require HTTP serving instead of data URL embedding for screenshots.
- Each screenshot uses a fresh browser instance (~40s per image) to avoid memory issues.
- Color mapping may differ slightly from ChimeraX rendering; information content is the same.
- Asymmetric large structures may render slightly off-center in some views due to a Mol* camera-framing quirk; structure and density remain visible.
- IHMValidation #127 Re-implement plotting routines with Mol*
- IHMValidation #119 Re-implement Q-score
- qscore-mapq Pure Python Q-score matching MapQ