Cell API

Read a segmentation table, optionally (but ideally) merged with a platemap.

• data_path — .csv or .parquet, a local path or an https:// URL. An unrecognized extension returns None; no exception is raised.

• platemap_path — CSV merged on Well. Unmatched wells are named in a warning.

• sample — draw a random subset of objects. Speeds up figure iteration on a large dataset.

• minutes_per_timepoint — derive a Time (min) column that every plot then uses as its x axis. See the caveat in the tutorial.

Returns a DataFrame.

Acquisition QC. One panel per metric: the object count, then the median of each channel in channels over time.

• channels — one channel name or a list of them, in the order the panels should populate. Either a channel ("Alexa Fluor 647") or the full column ("Intensity Mean (Alexa Fluor 647)") works.

• normalize — divide every channel by intensity at t=0.

Segmentation QC. A grid of segmented objects with mask outlines drawn on the image crops. Red is the selected object and teal are its neighbors.

pyramid_level should be set to the pyramid level segmentation was run at.

Median, IQR, and P10–P90 of value over time, per condition. This is the main population view.

Three rows: how many objects cross their well’s baseline gate, how bright those positives are, and how big each population is. Use it to tell whether more objects turned on or the same objects got brighter.

The gate is per-well — that well’s own gate_quantile of value at the first timepoint.

y against x as a hexbin, with a fit line in each panel. Answers whether size changes concentration. The fit covers all objects and is not split by percentile.

Empirical CDF at a few timepoints: what fraction of objects are dimmer than a given value. Shows the full distribution, where the quantile views show summary percentiles.

Walks an OME-NGFF plate zarr (wells → fields → timepoints), segments each frame with Cellpose, measures every channel, and appends the result to <dataset_name>.csv relative to the current working directory. If that CSV already exists it is renamed with a timestamp; nothing is overwritten.

segment_channel defaults to the first match among Rhodamine and Alexa Fluor 647. Both are membrane dyes, so segmentation runs on the membrane channel and not on the reporter. If neither is present, the function logs an error and skips the dataset.