Final Descriptors

The final descriptors stage recomputes descriptor values on the surviving molecule set near the end of the pipeline.

Purpose

This stage is primarily a reporting and comparison stage:

• it recalculates descriptors after all enabled filtering stages have narrowed the molecule set
• it gives the final report a clean “survivor profile”
• it supports side-by-side comparison between early and final descriptor distributions

Unlike the early descriptors stage, this step is usually interpreted as report enrichment rather than a primary screening gate.

Relationship to the Earlier Descriptors Stage

HEDGEHOG computes descriptors twice:

1. Initial descriptors to filter molecules early in the pipeline
2. Final descriptors to summarize the final survivors

Both runs use config_descriptors.yml, but they operate on different molecule populations.

Runtime Behavior

If no upstream data source exists for the final stage, or if no molecules remain, HEDGEHOG skips final descriptors rather than failing the whole run.

This is why the stage may be absent from a run that terminated early upstream.

Output

Final descriptor artifacts are written under:

stages/07_descriptors_final/

They feed directly into:

• report.html
• report_data.json
• final descriptor comparison plots in the report