Open Biomedical Network Benchmark | Continued development and maintenance

[arjunkrish]

Group

Krishnan Lab

Contact info

Arjun Krishnan, PI, [email protected], @Arjun Krishnan on slack.

Links to code

https://github.com/krishnanlab/obnb

Workflow

Please refer to https://github.com/krishnanlab/obnb/blob/main/README.md, which contains all this info.

Work description

This project concerns Open Biomedical Network Benchmark: A Python Toolkit for Benchmarking Datasets with Biomedical Networks https://proceedings.mlr.press/v240/liu24a.html

Open Biomedical Network Benchmark (OBNB) is a Python package that provides reusable modules for data downloading, processing, and split generation to set up machine learning (ML) on biological graphs (networks). It contains a collection of benchmarking datasets using publicly available biomedical networks (and gene annotation data) compatible with popular graph neural network frameworks such as PyG and DGL. It also contains implementations of several graph-based ML algorithms along with rigorous biologically meaningful evaluation schemes and metrics.

We wish to have the SET take on the continued development and maintenance of this package, working closely with members of the Krishnan lab. The first phase of this work will involve:

1. Applying software best practices and using input from our lab to refactor the code to make it easier to use.
2. Enabling users to add new networks, feature matrices, and prediction tasks.
3. Collaborate with members of our lab to improve the overall documentation and add tutorials (lab members will take the lead).

Timeline

There are no hard deadlines, but we would like to have the code refactoring (to improve usability) done in about 6 months, implementation of the new features (to add network, matrices, and tasks) done in the next 6 months. Throughout this time, members from our group will be closely involved in provided feedback and guidance on the changes, and developing documentation/tutorials.

Funding

We have some external support to do this work. Happy to increase our current arrangement with the SET.