Skip to content

README.md

Latest commit

 

History

History
184 lines (148 loc) · 7.29 KB

README.md

File metadata and controls

184 lines (148 loc) · 7.29 KB
Raw

Examples Overview

This directory contains the committed end-to-end reference examples for raw ColabFold-like outputs -> AFDB integration artifacts.

Recommended References

The two committed runnable references are:

Both references include:

  • normalized *-meta_v1.json + *-model_v1.pdb inputs
  • converted confidence and PAE JSONs
  • merged manifests
  • per-model and per-chain metadata JSONs plus batch files
  • ModelCIF input JSONs
  • generated ModelCIF files
  • DSSP-enriched mmCIF files
  • enriched PDB files
  • BCIF outputs with backend/fallback notes
  • exact regeneration commands in config/commands.txt

For the complex reference, the committed complexPredictionAccuracy_* fields in the JSON metadata are computed outputs from the iPSAE enrichment stage, not template literals. The ModelCIF template under uniprot/templates/colabfold_example_modelcif_metadata.json only supplies static ModelCIF scaffolding plus software parameter definitions; the computed complex metrics are injected from the enriched complex model JSONs before ModelCIF export.

The committed DuckDB subset examples/uniprot_example_subset.duckdb contains only the six UniProt accessions needed by the monomer and complex references, so the examples no longer depend on a VM-local UniProt database.

Regenerate the monomer reference from the repo root with:

.venv/bin/python scripts/generate_colabfold_e2e_example.py \
  --duckdb examples/uniprot_example_subset.duckdb \
  --output-dir examples/colabfold_monomer_e2e

Regenerate the complex reference with:

.venv/bin/python scripts/generate_colabfold_e2e_example.py \
  --duckdb examples/uniprot_example_subset.duckdb \
  --output-dir examples/colabfold_complex_e2e \
  --example-id AF-0000000066074510 \
  --example-id AF-0000000300000101

See examples/colabfold_monomer_e2e/README.md and examples/colabfold_complex_e2e/README.md for the selected fixtures, caveats, and generated file layout.

Validate Committed E2E Outputs

Run these checks from the repo root after installing the project environment. They validate one representative monomer and one representative complex from the committed reference trees.

Metadata JSONs

Use run-schema-validation with the dedicated example-output schemas for committed model_jsons/*.json, chain_jsons/*.json, and config/provider.json files, and use the same validator on the committed batch files:

.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_monomer_e2e/model_jsons/AF-0000000300000001.json \
  -t model-summary
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_monomer_e2e/model_batches/AF-metadata-1-of-1.json \
  -t model-summary
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_monomer_e2e/chain_jsons/AF-0000000300000001.json \
  -t collection-doc
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_monomer_e2e/chain_batches/AF-chain-metadata-1-of-1.json \
  -t collection-doc
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_monomer_e2e/config/provider.json \
  -t provider

.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_complex_e2e/model_jsons/AF-0000000300000101.json \
  -t model-summary
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_complex_e2e/model_batches/AF-metadata-1-of-1.json \
  -t model-summary
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_complex_e2e/chain_jsons/AF-0000000300000101.json \
  -t collection-doc
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_complex_e2e/chain_batches/AF-chain-metadata-1-of-1.json \
  -t collection-doc
.venv/bin/python main.py run-schema-validation \
  -i examples/colabfold_complex_e2e/config/provider.json \
  -t provider

For these committed e2e references, use model-summary for model_jsons/*.json and model_batches/*.json, collection-doc for chain_jsons/*.json and chain_batches/*.json, and provider for config/provider.json. The canonical model schema remains for full model metadata entries only.

The validate-metadata-file command uses the same shared metadata schema validator as run-schema-validation and requires the same explicit --type value, for example:

.venv/bin/python main.py validate-metadata-file \
  --file examples/colabfold_monomer_e2e/model_jsons/AF-0000000300000001.json \
  --type model-summary

Score JSONs

Validate the confidence JSON, the PAE JSON, and the confidence/PAE length relationship:

.venv/bin/python main.py validate-plddt-file \
  --file examples/colabfold_monomer_e2e/scores/AF-0000000300000001-confidence_v1.json
.venv/bin/python main.py validate-pae-file \
  --file examples/colabfold_monomer_e2e/scores/AF-0000000300000001-predicted_aligned_error_v1.json
.venv/bin/python main.py validate-relationships-pair \
  --plddt-file examples/colabfold_monomer_e2e/scores/AF-0000000300000001-confidence_v1.json \
  --pae-file examples/colabfold_monomer_e2e/scores/AF-0000000300000001-predicted_aligned_error_v1.json

.venv/bin/python main.py validate-plddt-file \
  --file examples/colabfold_complex_e2e/scores/AF-0000000300000101-confidence_v1.json
.venv/bin/python main.py validate-pae-file \
  --file examples/colabfold_complex_e2e/scores/AF-0000000300000101-predicted_aligned_error_v1.json
.venv/bin/python main.py validate-relationships-pair \
  --plddt-file examples/colabfold_complex_e2e/scores/AF-0000000300000101-confidence_v1.json \
  --pae-file examples/colabfold_complex_e2e/scores/AF-0000000300000101-predicted_aligned_error_v1.json

ModelCIF Dictionary Validation

Validate representative ModelCIF files with gemmi and the ModelCIF dictionary:

gemmi validate -p -d mmcif_ma.dic \
  examples/colabfold_monomer_e2e/modelcif/AF-0000000300000001-model_v1.cif
gemmi validate -p -d mmcif_ma.dic \
  examples/colabfold_complex_e2e/modelcif/AF-0000000300000101-model_v1.cif

Manual Coordinate-File Viewer Checks

Open representative PDB, ModelCIF, and BCIF files in the Mol* web viewer:

  1. Go to https://molstar.org/viewer/.
  2. Drag and drop the file into the browser window, or use Open Files in the left panel.
  3. Confirm the structure opens correctly, no error messages are shown in the viewer, and the structure looks structurally correct by eye.

Representative files:

examples/colabfold_monomer_e2e/modelpdb/AF-0000000300000001-model_v1.pdb
examples/colabfold_monomer_e2e/modelcif/AF-0000000300000001-model_v1.cif
examples/colabfold_monomer_e2e/bcif/AF-0000000300000001-model_v1.bcif

examples/colabfold_complex_e2e/modelpdb/AF-0000000300000101-model_v1.pdb
examples/colabfold_complex_e2e/modelcif/AF-0000000300000101-model_v1.cif
examples/colabfold_complex_e2e/bcif/AF-0000000300000101-model_v1.bcif

Optionally open the same files in ChimeraX or another preferred structure viewer such as PyMOL. The expected result is a clean import with no parser/import errors and a structure that looks correct by eye.