Pool replicate samples instead of refusing the run

[mzueva]

Greptile Summary

This PR replaces the duplicate-sample refusal with replicate pooling and exposes pooling details through the manifest, CLI report, and run-statistics UI. Important touched terms:

• Replicate sample — an additional sample sharing a condition-and-gate group; it is now pooled instead of causing refusal.
• Pooled group — a condition-and-gate group containing multiple samples; the new PooledGroup type records its condition, gate, sample IDs, and fraction disagreement.
• Sort fraction — normalized per-gate cell yield used for corrected scoring; replicate values are now averaged, although the current ordering can conceal invalid original values.
• Run manifest — the computation-to-model result contract; it now includes pooledGroups.
• Sample label — the user-facing name associated with a sample-axis ID; new model outputs let the statistics dialog resolve pooled sample IDs to labels.
• Read distribution — per-variant, per-gate display data; pooling restores one row per (variantKey, gate) for replicated gates.

Confidence Score: 4/5

The fraction-validation ordering should be fixed before merging because replicated inputs can conceal invalid per-sample values and produce corrected scores.

Replicate fractions are reduced to a non-null mean before the existing presence and range checks run, so an invalid original value can become an accepted pooled value.

Files Needing Attention: software/src/pooling.py, software/src/pipeline.py, software/src/validate.py

Important Files Changed

Filename	Overview
software/src/pooling.py	Adds replicate read and metadata aggregation, but averaging before validation can hide invalid per-sample fractions.
software/src/pipeline.py	Inserts pooling before validation and scoring and adds pooled-group details to the manifest.
software/src/validate.py	Removes duplicate-sample refusal and validates only the post-pooling fraction values.
model/src/types.ts	Adds the typed pooled-group manifest contract.
model/src/index.ts	Publishes the sample-label frame and column identifier used by the UI.
ui/src/components/RunStatistics.vue	Displays pooling warnings and resolves sample IDs to user-facing labels with a raw-ID fallback.
software/tests/unit/test_pooling.py	Thoroughly covers pooling grain, read sums, reporting, fraction averaging, and distribution uniqueness, but not invalid raw fractions masked by averaging.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Selected-gate read rows] --> B[Group replicate samples]
  B --> C[Sum reads by condition, gate, and variant]
  B --> D[Average replicate sort fractions]
  C --> E[Pooled reads table]
  D --> E
  E --> F[Validate pooled fractions]
  F --> G[Score retained conditions]
  G --> H[Score and distribution files]
  B --> I[pooledGroups report]
  H --> J[Run manifest]
  I --> J
  J --> K[CLI report and Run Statistics UI]

Loading

Prompt To Fix All With AI

### Issue 1
software/src/pooling.py:78
**Averaging bypasses fraction validation**

When a replicated gate contains an invalid sort fraction alongside a valid value, `pool_replicates` averages the raw values before validation; for example, `1.5` and `0.5` become the accepted value `1.0`. The subsequent presence and range checks therefore cannot identify the invalid sample, causing corrected scoring to proceed with input that should have been refused.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

_{Reviews (1): Last reviewed commit: "Pool replicate samples instead of refusi..." | Re-trigger Greptile}

Greptile also left 1 inline comment on this PR.

Context used (5)

• Knowledge Base — Sequence analysis execution
• Knowledge Base — Sequence input validation
• Knowledge Base — Analysis result I/O
• Knowledge Base — Analysis data model
• Knowledge Base — Analysis UI components and settings