# Synthetic supporting protocol — ST-301

**Status:** fabricated training material for the SHIRO AI analysis workflow. This document contains no real scientific, clinical, patient or proprietary information.

## 1. Purpose

Assess whether a fictional treatment series shows a different descriptive response from a matched control series over 0, 24, 48, 72 and 96 hours. The first-pass review should also identify data-quality issues before anyone interprets the pattern.

The intended analysis question is:

> Are treatment measurements higher than controls over time, and which records need review before interpreting the result?

This is a synthetic assay export created to exercise a source-aware analysis workflow. It is not evidence of efficacy, toxicity, mechanism or biological significance.

## 2. Study design

- **Study ID:** ST-301
- **Groups:** Control and Treatment
- **Timepoints:** 0, 24, 48, 72 and 96 hours
- **Batches:** B01 to B04
- **Replicates:** three recorded replicates per group and timepoint
- **Expected design:** 30 planned observations. The export contains one repeated physical row, so it has 31 data rows and should trigger duplicate review.
- **Primary comparison:** descriptive group and timepoint patterns only

The treatment and control rows are not paired biological samples. Do not infer causality or statistical significance from this export alone.

## 3. File and field definitions

The companion file is `shiro-st301-synthetic-study.csv`. It is a flat table with one observation per row and one variable per column.

| Field | Definition | Unit or allowed values |
|---|---|---|
| `Study_ID` | Study identifier | `ST-301` |
| `Sample_ID` | Unique sample and replicate identifier | Text; expected to be unique per observation |
| `Group` | Experimental group | `Control` or `Treatment` |
| `Batch` | Processing batch identifier | `B01` to `B04` |
| `Timepoint_h` | Collection time from baseline | Hours; expected values 0, 24, 48, 72, 96 |
| `Replicate` | Recorded replicate number within a timepoint and group | Positive integer 1 to 3 |
| `Concentration_mg_L` | Measured concentration | mg/L; expected to be present and greater than 0 |
| `Methylation_percent` | Methylation measurement | Percentage; expected range 0 to 100 |
| `Viability_percent` | Viability measurement | Percentage; expected range 0 to 100 |
| `Response_AU` | Instrument response signal | Arbitrary units; expected to be non-negative |
| `QC_Status` | Recorded source quality label | `PASS` or `REVIEW` |
| `Notes` | Source-facing observation note | Free text; do not use as a numeric measurement |

## 4. Review and validation rules

1. Preserve the uploaded values. Do not silently correct, delete, deduplicate or impute records.
2. Flag a missing required measurement, especially `Concentration_mg_L`.
3. Flag exact duplicate rows and check whether the repeated record is a file-generation or source-system issue.
4. Confirm that `Methylation_percent` and `Viability_percent` remain between 0 and 100.
5. Confirm that concentration is expressed in mg/L before comparing timepoints or batches.
6. Check that each group has the intended timepoints and replicate coverage.
7. Treat `QC_Status=REVIEW` as a review cue, not as proof that the measurement is invalid.
8. Keep the `Notes` field separate from measured values during numeric analysis.

## 5. Controlled review cues in this training export

This synthetic file intentionally contains three issues so the workflow has real review work to display:

- one blank concentration value at the Treatment, 72-hour, replicate 2 record;
- one methylation value of 104.2%, repeated in the duplicate row;
- one exact repeated Treatment, 72-hour, replicate 3 row.

These cues are deliberately visible in the source file and should be cited in any report. They must not be described as findings about a real experiment.

## 6. Interpretation boundaries

The dataset can support descriptive summaries and charts of the recorded values. It cannot support a claim of biological effect, clinical relevance, mechanism, safety, regulatory compliance or statistical significance without an approved analysis plan, validated methods, complete source records and qualified scientific review.

Before relying on any output, resolve the flagged records, confirm units and assay definitions, verify batch and replicate provenance, and repeat the analysis using the corrected source export. A qualified scientist remains responsible for interpretation and decisions.