Circe v0.4.0 — Refactoring and Performance Updates

New Features

• Rich progress bars — Graphical lasso chromosome processing now display Rich progress bars with integrated status messages.
• NumPy 2.x support

Refactoring

• Extracted graphical lasso functions into a dedicated gl_network module.
• Extracted private helpers in latent_network.py for improved readability.
• Centralized organism parameter resolution via a shared helper function.
• General code quality improvements and package modernization.

Bug Fixes

• Fixed SettingWithCopyWarning by adding .copy() calls where needed.
• Fixed isinstance check for AnnData in get_distances_regions.
• Fixed circular import in gl_network.py.
• Fixed chromosome ordering bug.
• Fixed sort_regions returning a view instead of a copy.
• Cleaned up redundant mask logic and unused parameters.
• Ensured progress bars render 100% on completion.
• Added build dependencies.

Contributors

Thank you @shahrozeabbas for the amazing refactoring!

Minor fixes - CIRCE 0.3.9

• Improving CIRCE speed to extract network links
• CIRCE now compatible with anndata<0.12
• Loosened dependency requirement for pandas

Full Changelog: v0.3.8...v0.3.9

1st release for handling huge datasets.

What's new:

• Parallelize choice of random windows and parameter calibration
• Updated memory usage and covariance to only store non only-0s cells per DNA window
  --> Work on atlases
• Accept chromosome sizes as a parameter (should very slightly change results if used)
• Minor bug fixes

Release for the 1st manuscript.

What's new:

Parallelize run on chromosomes
Speed up alpha calculation
Visualize gene coordinates alongside the DNA regions network