Quality control
Registration failure is the main risk in a cyclic-IF experiment, and it does not always announce itself. A misaligned plate still produces a full dataset and a plausible-looking plot — the numbers are simply from the wrong cells.
Run the QC report before trusting any 4i result.
python bin/alignment_qc.py 4126 -o alignment_qcPass several master plate IDs to treat them as biological replicates and add a combined, replicate-level analysis.
| Option | Default | Purpose |
|---|---|---|
-o, --outdir DIR |
alignment_qc |
Destination for CSVs, tables and figures |
--sample-alignments |
off | Must match how aggregate_plates was run |
--iqr N |
1.5 | IQR factor for per-well agreement outlier removal |
--mask-samples N |
12 | Fields sampled for nuclear-mask overlap (0 to skip) |
--pixel-size N |
— | Override the pixel size |
--attach |
— | Attach the report back to the OMERO plate |
--seed N |
random | Random seed for sampling |
--env NAME |
— | Environment file to load |
What the report contains
Everything for one run lands in a single directory: the raw input CSVs, the derived metric tables, summary and statistics tables, and the figures. That makes it archivable alongside the analysis — the evidence that a given dataset was correctly registered travels with the dataset.
--attach additionally uploads the report to the OMERO plate, which is the better choice when the analysis is shared.
Reading the result
The two questions the report answers are whether individual rounds agree with the reference, and whether that agreement is uniform across the plate. Registration that is good in the plate centre and poor at the edges points at a stage-position or stitching problem rather than a staining one.
--mask-samples controls how many fields are sampled for direct nuclear-mask overlap. It requires downloading masks, so it is the slow part of the report; set it to 0 for a quick pass and raise it when a plate looks suspect.