Skip to content

v0.9.4

Choose a tag to compare

@rabyj rabyj released this 10 Feb 22:41
· 77 commits to master since this release
v0.9.4
04045b2

Adds a general training interface, new utility scripts and furthers offline analysis support.

General Training Script

  • Added general_training.py: Simplified script to train MLP classifiers without EpiATLAS metadata (ignore uuid/EpiRR).
  • Added a Makefile for a more standard interface to run tests
  • Run make demo-test from the test folder to see general training script in action.
  • Added an integration test for the new training workflow.

Offline Analysis Support

  • Analysis class now works without an experiment logger (logger=None).
  • New save_dir parameter allows saving predictions and confusion matrices locally.
  • Added safeguards to skip logging calls when no logger is present.
  • Fixed a potential undefined variable bug in _generic_write_prediction.

New Notebooks & Analysis

  • Added notebook comparing MLP recount3 sex prediction metrics with naive chrY/chrX read count classifiers.
  • Added notebook comparing GO enrichment of SHAP-selected regions against effect-size baselines with matched region counts.

New Utilities

  • compress_hdf5_into_npz.py: converts HDF5 files to compressed NPZ format for faster data loading.
  • bins_to_bed: new bed file utility.
  • Made HDF5 filepath adaptation optional.

Developer Experience

  • Improved training and prediction script templates with better file/folder existence checks.
  • Moved install.py to repository root for better accessibility; made it robust with recursive pyproject.toml search and cross-platform compatibility.
  • Added pytest-xdist for parallel test execution and introduced a slow test marker to enable easier exclusion (make test-fast uses it).
  • Added compressed fixtures index to track binary test data state.
  • Updated requirements: generated from latest py3.10-3.12 test suite.

Full Changelog: v0.9.3...v0.9.4