Skip to content

Releases

ProtCell v1.0.0: pretrained model and example data

Choose a tag to compare

@hubwenbwen hubwenbwen released this 11 Oct 08:33

ProtCell v1.0.0

This release provides the pretrained ProtCell checkpoint and processed inputs for the five downstream notebooks.

Assets

  • protcell_example_data.tar.gz: the results/data/ directory, including the study datasets, fixed Sparks2023 discovery/selection/test splits, discovery statistics and panel definitions.
  • protcell_pretrained.pt: the pretrained checkpoint loaded by the notebook setup cells.
  • protcell_pretrained.json: checkpoint provenance and SHA-256 checksum.
  • SHA256SUMS: checksums of the data archive, checkpoint and checkpoint metadata.

Use

Extract the data archive from the repository root:

tar -xzf protcell_example_data.tar.gz
mkdir -p checkpoints
mv protcell_pretrained.pt protcell_pretrained.json checkpoints/

Follow the repository README to install the environment, then run the notebooks in results/. The data bundle contains already-preprocessed example inputs. results/data/manifest.json lists the individual files, studies, matrix shapes and checksums.