Skip to content

Version 3.1.0

Choose a tag to compare

View all tags
@TsingYang1112 TsingYang1112 released this 04 Aug 15:21
· 16 commits to main since this release
v3.1.0
d416230

Highlights

  • CV threshold auto-search: --cv_rank_thresh now supports single values, comma-separated lists, ranges (e.g., 0.3-0.7:0.1), and auto mode.
  • Optimal threshold selection: Automatically selects threshold that minimizes Ω (pre-QC) + Ω (final).
  • Organized output structure: Results for each CV value stored in separate directories; best results copied to 05_final_results/.
  • New module: full_tree_builder.py encapsulates Steps 6-8.
  • Spectraformer mode: Complete support for spectraformer workflow integration.

What's New

Core Algorithm

  • CV threshold auto-search with auto mode for optimal selection.

New Features

  • full_tree_builder.py module for streamlined Steps 6-8.
  • Spectraformer workflow integration.

Bug Fixes

  • Tree-Matrix consistency: Fixed on_node placement to keep tree and matrix synchronized; resolves KeyError.
  • Multi-mutation node handling: Fixed column name mismatch when multiple mutations (e.g., mutA|mutB) are consolidated into a single node in T_current but remained as separate columns in M_current; now M_current columns are named consistently with node labels.
  • Fixed cluster_external_mutations_by_intersection import error.
  • Fixed high_cv_mutations NameError in scaffold building.
  • Fixed mutation reattachment logic bugs.

Performance

  • Numpy-vectorized penalty scoring.
  • Reduced deep-copy overhead in Bayesian scoring.
  • Accelerated Leiden graph clustering.

Code Quality

  • Code refactoring with improved annotations.
  • Optimized logging and comments.
  • Removed unnecessary intermediate files.

Notes

Assets 2
Loading