# 4i / Cyclic IF

Cyclic immunofluorescence images the same plate several times, stripping and re-staining between rounds. Instead of the three or four markers a single round allows, you measure dozens per cell -- provided the rounds can be registered accurately enough to say that a nucleus in round 3 is the *same* nucleus as in round 1.

That registration problem is what this part of OMERO-Screen solves.


# The idea

Each round is analysed as an ordinary plate: segmented, measured, stored. The 4i tools then run afterwards, computing a geometric alignment between rounds, matching cells across them, and merging the per-round features into one single-cell table.

Nothing about the per-round analysis is special. If you can run a plate, you can run a 4i experiment.


# The workflow

| Step | Tool | What it does |
|----|----|----|
| 1 | `omero-screen` | Analyse **each round** as a normal plate |
| 2 | [`align_plates.py`](alignment.md) | Register the rounds against a reference |
| 3 | [`create_masks.py`](alignment.md#filling-in-missing-masks) | Fill in cell masks missing from secondary rounds |
| 4 | [`aggregate_plates.py`](aggregation.md) | Merge per-cell features across rounds |
| 5 | [`alignment_qc.py`](quality-control.md) | Check the registration actually worked |

The first plate passed to `align_plates.py` is the **reference**. Every other round is registered onto it, and the aggregated dataset carries its plate ID.


# Before you start

> **Warning: Warning**
>
> These are `bin/` utilities and still use argparse, unlike the four main commands. They keep argparse conventions -- including abbreviated options -- and are run with `python bin/<name>.py` rather than as installed commands.

> **Important: Important**
>
> `--sample-alignments` must be set **consistently** across steps 2, 3 and 4. Mixing per-sample and per-well alignments produces silently wrong mappings -- no error, just cells matched to the wrong partners.


# Analysing the rounds

Run the pipeline on each round separately, exactly as for a single-round experiment:

``` bash
uv run omero-screen 4126 --stitch
uv run omero-screen 4127 --stitch
uv run omero-screen 4128 --stitch
```

Each round gets its own segmentation, features and QC. At this point nothing knows the rounds are related -- that begins with [alignment](alignment.md).


# Protocols

> **Note: Note**
>
> Wet-lab protocols for the stripping and re-staining rounds are being prepared and will be published here.