v0.4.0

v0.4.0

• annotate_hmm.py:

  • Changed parallelization from multiple concurrent processes (1 CPU each) to a single process using all available CPUs via pyhmmer's native threading.
  • Now splits input sequences into chunks only for very large inputs.
  • Reduces memory usage and improves speed for large datasets; similar runtime for smaller datasets.

• filter_by_length.py, filter_by_cds.py:

  • Revised parallelization to prevent bottlenecks from a few large sequences slowing all jobs.

• check_circular.py:

  • Improved batching and parallelization, drastically reducing runtime when there are some very long sequences.

• genome_context.py, curate_annots.py:

  • Removed max_flank_length enforcement; no longer report binary flanking gene flags.
  • Instead, now report exact distances to the nearest V-score=10, viral hallmark, or MGE gene as features.
  • Stopped imputing missing features since LGBM can handle np.nan directly.

• main.py, CheckAMG_annotate.py, CheckAMG_annotate.smk, organize_proteins.py, make_final_table.py:

  • Minor compatibility updates for the above changes.

• Re-trained the viral origin confidence LGBM on a more robust dataset

  • Updated the model .joblib files for the new model
  • Updated the precision-recall curve plot and table