Host–Pathogen Analysis¶

From Home → Assays → Toxoplasma, open Host–Pathogen Analysis. This alpha module combines vacuole-level marker recruitment with host and well summaries. Existing Recruitment remains available.

Try the real microscopy test data¶

Open Home → Assays → Toxoplasma → Host–Pathogen Analysis and select Load test data…. The download is approximately 114 MB.
Enable Live beside the bottom action controls, then select Run preview. Choose either measured field and select vacuoles in the image or table to inspect their host links and marker ratios.
Set Image channel to 2 for Toxoplasma, 1 for RNF213 or 3 for CellMask. Leave the mask-plane selectors on Auto; the dataset includes the plane manifest. These display choices do not change the analysis channels.
Turn Live off and expand Actions if the run controls are collapsed. Select Run to analyze both fields. Compare the results with the supplied example_cells.csv, example_vacuoles.csv and example_wells.csv.

Follow the Host–Pathogen video walkthrough for the complete example.

The Host–Pathogen dataset contains two acquired THP-1/RNF213 fields, four intensity channels and prepared automatic masks. Its measurements cover 164 host cells, 189 nuclei, 97 whole vacuoles and 164 cytoplasm objects across the two fields. The acquired intensity planes are unchanged; the masks underwent Measure’s normal parent/child reconciliation so the distributed outlines match measured objects. The dataset card and example_manifest.json describe the source fields, preparation and checks. These masks are not manually validated ground truth.

The supplied settings compare vacuole RNF213 means with their associated host cytoplasm means in channel 1. Marker thresholds are deliberately unset: ratios are available, while positive/negative marker states remain unknown until you choose suitable control-calibrated thresholds. Individual-parasite counts are not supplied, so replication remains not_measured. This small sample does not establish biological differences between its control conditions.

The command-line download is spacr-download host_pathogen. The prepared measurements let you use this example directly; your own image project needs the Mask and Measure preparation described below.

Inspect one field before running¶

After selecting a measured project in Source, enable the bottom Live control. The preview uses the current form settings and the same analysis functions as a full run. Select Run preview for the initial calculation. While the panel remains visible, later settings changes refresh that preview. Use Refresh to reload the field choices and Cancel to cancel a pending preview. Preview reads the measurements database without writing result files.

Its host counts and infection fractions describe the selected field only. They are not whole-well denominators. Fields containing vacuoles without measured host cells remain available. Missing or zero reference intensities produce unknown marker states; an unknown state must not be read as negative.

Overlay plane selection defaults to Auto, using the recorded .spacr_plane_layout.json metadata. Select explicit host, vacuole and parasite plane indices when metadata is absent, or use -1 to hide an overlay. Changing the displayed intensity channel does not change the analysis marker settings. If an image cannot be loaded, the measured results remain available in the table.

For scripted previews, spacr.host_pathogen_preview.preview_fields() offers at most 50 fields by default and reports whether more exist. spacr.host_pathogen_preview.preview_field() analyzes one selected field, refusing input tables with more than 100,000 rows for that field by default. These bounds keep the preview limited; use the full analysis for project-wide reports.

Prepare the counting units¶

Run Mask and Measure first. Keep uninfected host cells in Measure (uninfected=True) when the intended infection denominator is all measured hosts. Previously discarded host cells cannot be reconstructed here.

Supply a cell table, a table with one object per whole vacuole (default: pathogen), and a host reference compartment table (default: cytoplasm). Individual-parasite masks cannot substitute for whole-vacuole masks. Select marker channels and per-channel ratio thresholds calibrated with appropriate controls; the module does not learn these thresholds.

Replication counts require either a parasite table with explicit parent-vacuole identities or an existing count column. Choose one count source. Without one, replication remains unmeasured rather than being inferred from vacuole area. Host identity alone is insufficient when a host contains multiple vacuoles.

When Measure links individual organelle-role objects to whole-vacuole masks, one parent must cover strictly more than half of a child’s pixels. Outside objects and ambiguous overlaps retain a missing parent. The stored parent is pathogen_id; the stored overlap feature is role-prefixed, for example organelle_pathogen_overlap_fraction. Use such an object table as parasites only when those masks actually represent individual parasites.

Interpret the reports¶

The output directory results/host_pathogen contains vacuoles.csv, cells.csv, wells.csv, marker_states.csv, replication_distribution.csv, orphan_parasites.csv and the exact settings.json. With multiple input projects, the combined report is saved under the first project. Database paths separate sources even when plate or field identifiers repeat.

Unlinked vacuoles and missing, invalid or zero host-reference intensities yield unknown marker states. Joint marker fractions include unknown vacuoles in their denominator; replication fractions use only vacuoles with counts. Host infection fractions describe retained measured host cells. These denominators answer different questions and should be reported explicitly.

For scripted use, see spacr.host_pathogen.analyze_host_pathogen() and spacr.host_pathogen.summarize_tables(). The headless module name is host_pathogen in spacr-run.

Choosing a Replication Assay method¶

The separate Replication Assay → Method selector defaults to direct_count, which counts individually segmented parasites per assigned vacuole. size_proxy uses the existing endodyogeny analysis and its Size Proxy (Legacy) controls. That readout aggregates pathogen area per host cell; multiple vacuoles in one host are combined. It is an area-derived proxy, not a parasite count or measured volume.

Both available routes return replication_method metadata. The whole-vacuole deep-learning option is marked coming soon and cannot run before its model is available. See spacr.submodules.analyze_replication() and spacr.submodules.analyze_endodyogeny() for the exact settings and outputs.