Skip to content

Roadmap

This roadmap is intentionally conservative. MolScope is useful today as a lightweight teaching, prototyping, and ML-representation toolkit; future work should strengthen that identity rather than turn it into a full simulation framework.

Product focus

New work should make at least one of these workflows noticeably better:

  • PDB to descriptors.
  • PDB to graph/GNN.
  • PDB to coarse-grained beads.

Features outside those paths should stay experimental, optional, or documented as supporting capabilities until they earn a clearer role.

v0.9

  • Improve the documentation site structure and API reference.
  • Polish the three core tutorials and keep examples aligned with them.
  • Expand benchmark coverage for parsing, contact maps, graph export, and descriptor generation.
  • Grow scientific validation from the current 1fqy/small-molecule checks into a curated mini-panel for DSSP, geometry, ensembles, and bond perception.
  • Make validation results easier to inspect from CI logs and docs.

v1.0

  • Declare a stable core API for Molecule, readers/writers, descriptors, contact maps, and graph export.
  • Freeze descriptor and graph feature preset names where practical.
  • Clarify deprecation policy for old helper APIs and compatibility shims.
  • Publish a concise migration guide from pre-1.0 versions.

Future

  • Trajectory-lite support for small multi-frame XYZ/PDB workflows.
  • Better CIF/mmCIF coverage while keeping Gemmi optional.
  • More configurable graph edge construction for ML workflows.
  • More explicit coarse-grained topology objects and export formats for prototyping.
  • Optional generated API documentation from docstrings.

Recently resolved

  • PyPI publishing with trusted release workflow.
  • CITATION.cff citation metadata.
  • ML tutorial from PDB ensemble to PyTorch Geometric graph-level learning.
  • Optional Gemmi-backed mmCIF syntax, atom-site, and dictionary validation hooks.
  • Stable descriptor presets for native structural and RDKit-backed feature sets.
  • Graph node/edge featurization presets for ML workflows.
  • Residue-level contact graphs with NetworkX, PyG and DGL exporters.
  • Convenience extras for NetworkX, PyTorch Geometric, DGL, RDKit, and Gemmi.
  • Chunked distance histograms and contact-count paths for larger structures.