Skip to content

TargetIntel-IO v0.1.3 — End-to-end CLI and portfolio refresh

Choose a tag to compare

View all tags
@rsolerortuno rsolerortuno released this 10 Jul 11:38
v0.1.3
617a8ae

TargetIntel-IO v0.1.3

This release improves the usability and portfolio presentation of TargetIntel-IO by introducing a complete command-line workflow.

End-to-end command-line interface

The full target-triage workflow can now be executed with:

targetintel run

This generates:

  • the processed target feature table;
  • therapeutic-intent-aware rankings;
  • Markdown hypothesis cards;
  • browsable HTML reports;
  • summary figures and rank-shift visualizations.

The complete internal validation workflow is available through:

targetintel run --validate

This additionally runs:

  • the curated 56-target benchmark;
  • the augmented 331-target benchmark universe;
  • the 42-scenario scoring-weight sensitivity analysis;
  • the annotated sensitivity heatmap.

Additional options include:

  • targetintel run --refresh
  • targetintel run --help

Portfolio and documentation improvements

  • Reorganized the main README around the problem, outputs, quick start, CLI, validation results, limitations, and reproducibility.
  • Removed duplicated and obsolete prototype documentation.
  • Added direct links to versioned reports, figures, benchmark outputs, and sensitivity results.
  • Added README regression tests.

Repository cleanup

Removed obsolete prototype artifacts:

  • scripts/01_target_prioritization_mvp.py
  • scripts/02_plot_top_targets.py
  • figures/test.png

The reusable stage-specific scripts and all versioned figures under examples/figures/ remain available.

Validation status

Current internal benchmark results:

  • 56/56 benchmark targets evaluated;
  • 91.07% strict primary-intent accuracy;
  • 100% acceptable-intent accuracy;
  • 90.57% cross-intent specificity;
  • 58.10% mean top-10 recall;
  • 79.46% mean top-20 recall.

The repository includes 92 passing unit, integration, and regression tests.

Scope

TargetIntel-IO remains an internal rule-based hypothesis-generation and target-triage framework for anti-PD-1-resistant melanoma. Its benchmark is internally curated and does not constitute independent biological or clinical validation.
EOF

Assets 2
Loading