v1.0.2

Addressed reported issues with v5/v6 datasets and released v7 datasets:

• Propagated Nunique column from HybriDetector output to the final datasets
• Propagated all candidate miRNA names for a given miRNA sequence in the noncodingRNA_name column from HybriDetector to prevent miRNA seqs mapping to multiple miRNA names and vice versa
• Generating negatives
  • Fixed off-by-one bug in sorting step
  • Clustered unique target sites (rather than all target sites)
  • Performed clustering cutoff selection analysis
  • Fixed mixed data types bug causing only partial cluster exclusion in generating candidate negatives step
  • Fixed method to drop (randomly) excess positive examples for miRNA families for which insufficient negative candidates are available (the code was raising not handling before); this helps maintain a positive-to-negative ratio closer to 1:1 
  • Refactored code to reuse functions and avoid duplication
• Re-annotated target sites using genomic_region_annotator tool, which adds the columns:
  • dominant_region: region with greatest overlap in the selected transcript
  • regions_present: all overlapping regions in the selected transcript
  • read_start_in_sel_tx_1based: 1-based transcript-relative start coordinate
  • read_end_in_sel_tx_1based: 1-based transcript-relative end coordinate
• Fixed all coordinate columns to integer data type
• Added a master post-process pipeline script for easier execution 
• Fixed potential train-val leakage by using a shared split for data generators in the training script
• Updated the relevant documentation

miRBench_paper v1.0.1

Added gene clustering analysis and engineered dataset with inflated miRNA frequency class bias, to address reviewer comments for miRBench manuscript submitted to ECCB/ISCB 2025 Proceedings (Bioinformatics).

miRBench_paper v1.0.0

Code for model training, miraw analysis, and new plots added.

miRBench manuscript submitted to ECCB/ISCB 2025 Proceedings (Bioinformatics) with this version of main branch.

miRBench_paper v0.2.0

Code refactored and new scripts added to address and mitigate against a bias discovered in previous versions of the datasets.

This release contains code used to create the reproducible, unbiased (miRNA families balanced between positive and negative classes) datasets available as v5 on Zenodo.

miRBench_paper v0.1.2

miRBench manuscript submitted to BioRxiv with this version of main branch and the following commits on yet unmerged branches (pending approval of PRs).

Branch david/fig5, commit 87a028f

Branch eva/model_training, commit cb4abfe

miRBench_paper v0.1.1

Updated to create a reproducible post-process pipeline, retaining also 'chr', 'start;', 'end', and 'strand' columns from the concatenated HybriDetector output files, to enable addition of other features downstream, such as conservation.

The datasets on miRBench 1.0.0 were created by this release.

miRBench_paper v0.1.0

First version for miRBench manuscript.