Hugging Face's logo

Datasets:
DataQuests
/
Dyna_Repo_Inria_MDPosit_Files

Tasks:
Token Classification
Modalities:
Image
Formats:
imagefolder
Size:
10K<n<100K
Tags:
chemistry
biology
medical
Synthetic
Libraries:
Datasets
License:
Dataset card Data Studio Files
xet
Community
1
Dyna_Repo_Inria_MDPosit_Files / README.md
Tonic's picture
Tonic
Update README.md
834a6f2 verified
preview code
|
raw
history blame contribute delete
11.7 kB
metadata
configs:
  - config_name: default
    data_files:
      - split: train
        path:
          - data/files/*/*screenshot*.jpg
          - data/files/*/screenshot.jpg
          - data/files/*/clusters_*_screenshot_*.jpg
          - data/files/*/*.jpg
          - data/files/*/*.jpeg
          - data/files/*/*.png
  - config_name: manifests
    data_files:
      - split: train
        path:
          - data/files/*/records.jsonl
          - data/files/*/project.json
          - data/files/all_records.jsonl
  - config_name: structures
    data_files:
      - split: train
        path:
          - data/files/*/*.pdb
          - data/files/*/*.tpr
  - config_name: trajectories
    data_files:
      - split: train
        path:
          - data/files/*/*.xtc
          - data/files/*/*.trr
          - data/files/*/*.dcd
          - data/files/*/*.bin
          - data/files/*/pca_trajectory_*.bin
license: gpl
task_categories:
  - token-classification
tags:
  - chemistry
  - biology
  - medical
  - synthetic
size_categories:
  - 10K<n<100K

DynaRepo MDPosit

This repository mirrors Molecular Dynamics (MD) project metadata and files laid out per accession as exposed by Dynarepo. It includes global and per-accession manifests, project-level JSON, and the actual files (structures, trajectories, derived PCA binaries, and analysis screenshots). Field names and shapes follow Dynarepo’s documented model.

Overview

How to consume

  • Stream-parse: Read the file line-by-line and json.loads each object to enumerate replicas and their artifacts.
  • Relate to project_example.json: Each line mirrors an item in flattened_md_records and can be joined back to project-level fields if needed.
  • Concatenate across accessions: You can build a global manifest by concatenating multiple data\files\<ACCESSION>\records.jsonl files.
  • Field semantics: Follow the Dynarepo docs: Meta, Statistics, Help.

Understanding all_records.jsonl (global flattened per-accession manifest)

  • Source: Global, line-delimited aggregation of per-replica records across all accessions. See: Dynarepo Meta, Dynarepo Statistics, and Dynarepo Help.
  • Where it lives: data\files\all_records.jsonl.
  • What it is: A JSON Lines file where each line mirrors a single replica entry from per-accession records.jsonl (fields include project_identifier, accession, md_index, md_name, analyses, metadata, and files).
  • How to use: Stream-parse line by line, filter by accession or analyses, and group by accession/md_index to enumerate all replicas and their artifacts across the dataset. Field semantics are identical to per-accession records and follow Dynarepo docs.

Remote folder contents in data\files\<ACCESSION> (A0098 example)

  • Source: Mirrors the per-accession file repository exposed by Dynarepo. See: Dynarepo Meta, Dynarepo Statistics, and Dynarepo Help.
  • Where it appears: Under data\files\<ACCESSION>\ in this repository; corresponds to Dynarepo’s remote path /projects/<ACCESSION>/files.
  • What you will typically find (A0098 shown):
    • Core descriptors: project.json (5.18 kB), records.jsonl (43.3 kB)
    • Core MD simulation assets: topology.tpr (1.71 MB), structure.pdb (397 kB), trajectory.xtc (93.5 MB), trajectory.bin (294 MB)
    • PCA-derived binaries: pca_trajectory_01.bin … pca_trajectory_05.bin (5 files, ~237 kB each)
    • Cluster analysis screenshots:
      • clusters_00_screenshot_00.jpg … clusters_00_screenshot_20.jpg (21 images, ~16–21 kB each)
      • clusters_01_screenshot_00.jpg … clusters_01_screenshot_19.jpg (20 images, ~16–21 kB each)
    • Cover/preview: screenshot.jpg (~20.7 kB)
  • Relationship to manifests:
    • Every file present under data\files\<ACCESSION>\ is enumerated in that accession’s records.jsonl files[] entries with name, downloadUrl (relative to /projects/<ACCESSION>/files/<name>), length (bytes), and contentType.
    • The same inventory is summarized in project.json under its files section (e.g., list, filenotes, merged) and joined into flattened_md_records for per-replica views.
  • Notes: Exact counts and sizes vary by accession and by which analyses were run; additional analysis types will add further artifacts alongside those listed above.

Understanding project.json (Dynarepo dataset A0098)

  • Source: Dataset metadata follows the Dynarepo model. See: Dynarepo Meta, Dynarepo Statistics, and Dynarepo Help.
  • Where it appears: Each dataset under data\files\<ACCESSION> includes a project-level JSON with this schema. This repository provides a canonical example in project_example.json.
  • What it is: A single, self-contained description of an MD project combining high-level project metadata, per-replica MD records, file inventory, topology, references, analysis results, and summaries. It is designed to be directly consumable by Dynarepo services and UIs.

High-level structure

  • project: Core project info and global metadata
    • Keys: accession, published, metadata (e.g., NAME, DESCRIPTION, AUTHORS, PROGRAM, VERSION, TYPE, METHOD, LICENSE, PDBIDS, REFERENCES, simulation params like FRAMESTEP, TIMESTEP, TEMP, ENSEMBLE, FF, WAT, box sizes, system counts), mds (replica names), analyses (available analysis types), files (canonical file names), mdIndex/mdNumber/refframe.
  • flattened_md_records: One entry per MD replica with the replica’s metadata and a resolved files list including downloadUrl, length, and contentType.
  • files:
    • list: canonical file names present in the project
    • filenotes: storage-level details (size, content type, upload date)
    • merged: convenient join of name, URL, size, and metadata
  • topology: Atoms, residues, chains, bonding, and external references used to build the system.
  • references: Protein and PDB references (e.g., UniProt IDs, sequences, domains) and cross-links to PDB entries.
  • analyses: Machine-readable outputs per analysis (e.g., clusters, dist-perres, energies, fluctuation, hbonds, interactions, pca, pockets, rgyr, rmsd*, sasa, tmscores).
  • summaries: Compact keys/metrics for quick inspection and validation.

How this maps to datapoint A0098

  • Accession: A0098 (see data\files\A0098\records.jsonl for the per-replica lines used to populate flattened_md_records).
  • Project metadata examples (from A0098):
    • metadata.NAME: "Dynarepo 5FZT"
    • metadata.PDBIDS: ["5FZT"]
    • metadata.REFERENCES: ["P26039", "Q96QB1"]
    • Simulation constants: FRAMESTEP: 0.1, TIMESTEP: 2, TEMP: 310, ENSEMBLE: "NPT", FF: ["CHARMM36m"], WAT: "TIP3P", BOXTYPE: "Dodecahedron"
    • System scale: mdFrames: 5001, mdAtoms: 4900, PROTRES: 329, SYSTATS: 4900
  • Files commonly present: topology.tpr, structure.pdb, trajectory.xtc, trajectory.bin, pca_trajectory_*.bin, and analysis screenshots (e.g., clusters_*_screenshot_*.jpg).
  • Analyses present: clusters, clusters-00, clusters-01, dist-perres, energies, fluctuation, hbonds, interactions, pca, rgyr, rmsd-pairwise, rmsd-perres, rmsds, sasa, tmscores.

Understanding records.jsonl (per-replica manifest in data\files\<ACCESSION>)

  • Source: Per-MD replica records consistent with Dynarepo’s flattened MD model. See: Dynarepo Meta, Dynarepo Statistics, and Dynarepo Help.
  • Where it appears: For each accession, a records.jsonl lives under data\files\<ACCESSION>\. This repository includes data\files\A0098\records.jsonl. A compact illustration is also provided at the repository root as records._examplejsonl.
  • What it is: A JSON Lines file; each line describes one MD replica and its files.

Record shape (keys you will see per line)

  • Top-level identifiers: project_identifier, accession, published
  • Replica info: md_index (0-based), md_name (e.g., "replica 1"), chains (if applicable)
  • Analyses available: analyses (e.g., clusters, rmsds, pca, sasa, ...)
  • metadata object: Biology and simulation parameters used by Dynarepo UIs and services. Common fields include NAME, DESCRIPTION, AUTHORS, GROUPS, CONTACT, PROGRAM, VERSION, TYPE, METHOD, LICENSE (and LINKCENSE), PDBIDS, REFERENCES, sequences/domains (PROTSEQ, NUCLSEQ, DOMAINS), simulation setup (FRAMESTEP, TIMESTEP, TEMP, ENSEMBLE, FF, WAT, BOXTYPE), system sizes (SYSTATS, SYSTRES, PROTATS, PROTRES, ...), optional INTERACTIONS, box sizes (BOXSIZEX/Y/Z), and per-replica summaries like mdAtoms and mdFrames.
  • files array: One entry per stored artifact with name, downloadUrl (relative to /projects/<ACCESSION>/files/<name>), length (bytes), contentType, and a small metadata dict (e.g., {project: <id>, md: <index>} or per-file stats for derived binaries).

Typical files for A0098 include topology.tpr, structure.pdb, trajectory.xtc, trajectory.bin, pca_trajectory_*.bin, and analysis screenshots like clusters_*_screenshot_*.jpg.

Example (A0098, truncated line 1) 

{
  "project_identifier": "A0098",
  "accession": "A0098",
  "published": true,
  "md_index": 0,
  "md_name": "replica 1",
  "analyses": ["clusters", "pca", "rmsds", "sasa", "..."],
  "metadata": {
    "NAME": "Dynarepo 5FZT",
    "PDBIDS": ["5FZT"],
    "FRAMESTEP": 0.1,
    "TIMESTEP": 2,
    "TEMP": 310,
    "ENSEMBLE": "NPT",
    "FF": ["CHARMM36m"],
    "WAT": "TIP3P",
    "mdAtoms": 4900,
    "mdFrames": 5001
  },
  "files": [
    {"name": "topology.tpr", "downloadUrl": "/projects/A0098/files/topology.tpr", "length": 1712912, "contentType": "application/octet-stream"},
    {"name": "structure.pdb", "downloadUrl": "/projects/A0098/files/structure.pdb", "length": 396935, "contentType": "chemical/x-pdb"},
    {"name": "trajectory.xtc", "downloadUrl": "/projects/A0098/files/trajectory.xtc", "length": 93513296, "contentType": "application/octet-stream"}
    // ... more files omitted ...
  ]
}

Contributing

When creating a new project JSON for another accession, minimally review/update:

  • Identifiers: project.accession, project.identifier, and metadata.NAME
  • Biology: metadata.PDBIDS, metadata.REFERENCES (UniProt), sequences/domains in references
  • Simulation: FRAMESTEP, TIMESTEP, TEMP, ENSEMBLE, FF, WAT, box sizes, atom/residue counts
  • Replicas: mds list and flattened_md_records[*].md_index/md_name
  • Files: Ensure files.list, files.merged, and each replica’s files[] match your storage paths and sizes
  • Analyses: Keep only the analyses you actually produced and include their data payloads

For reference on field meanings and expected shapes, consult the Dynarepo documentation: Meta, Statistics, Help.

References: Dynarepo Meta, Dynarepo Statistics, Dynarepo Help