TileScope-results

Analysis scripts, figure builders and tests for the TileScope paper. The scenario generator is independent of the core analysis engine.

What's where, by paper figure

Figures	Scripts under analyses/figures/scripts/	Numerical results
Figure 1: method overview	overview/	Supplementary Data 1
Figure 2, S1–S3: benchmark and comparator methods	benchmark/	Supplementary Data 1
Figure 3, S4–S6: human glioblastoma	greenwald/	Supplementary Data 2
S7–S8: Xenium	xenium/	Supplementary Data 3
Figure 4, S9–S10: mouse glioblastoma	mouse_publication/	Supplementary Data 4

Figure 2C comparator values are in Supplementary Data 1 at comparator_methods/comparator_values.csv; the accompanying comparator_methods_schema.csv defines the statistics and directions.

Data download

Zenodo: 10.5281/zenodo.23535235

Data are not included in this Git repository. Five separate archives are prepared for the deposit:

Download	Contents	Extract into
input_data.zip	Prepared spot/cell values, coordinates, ROI labels, gene definitions and run settings	TileScope-data/
Supplementary_Data_1.zip	Benchmark, comparator and overview measurements	TileScope-data/supplementary_data/
Supplementary_Data_2.zip	Human GBM measurements and selected spatial geometry	TileScope-data/supplementary_data/
Supplementary_Data_3.zip	All 17 Xenium samples: measurements, coverage and illustrated geometry	TileScope-data/supplementary_data/
Supplementary_Data_4.zip	Mouse all-pairs measurements and signature-selection evidence	TileScope-data/supplementary_data/

Each ZIP contains its named directory, a README and a relative file index with checksums. Input data are stored once, not copied into each supplementary archive.

• Read results: download the relevant Supplementary Data archive; no code needed.
• Redraw figures: download Input Data and the relevant Supplementary Data archive, then use the matching code versions.
• Rerun TileScope: use the prepared inputs and recorded settings in Input Data. Reconstructing those inputs from raw expression requires the original-study data described in Data access. The CSV API (run_tilescope_from_csv) and CLI take one pair per input file, with columns x, y, gene1, gene2, roi_id. The human and mouse score banks in input_data/{greenwald,mouse}/scores/<sample>.csv.gz carry one row per spot and one column per scored signature; to drive core, select the two columns of interest, rename them to gene1 and gene2, and pass the real signature names as feature_a_id / feature_b_id in Python or --feature-a-id / --feature-b-id on the CLI. Per-pair CSVs are already materialised for Xenium (input_data/xenium/cells/<Study>/<sample>/<pair>.csv.gz) and benchmark (input_data/benchmark/corpus/<scenario>/<case>/<role>.csv). The human and mouse runners perform this selection from --input-dir and create results under --output-dir; no pre-existing CurrentResults/ directory is needed.

Figure 1 is a schematic method illustration. Its synthetic input and the retained plotting measurements under Supplementary_Data_1/overview/ are sufficient to re-render the published panels; a complete original TileScope run of this synthetic input is not required.

Install

git clone https://github.com/cbib/TileScope-results.git
cd TileScope-results
python -m venv venv
source venv/bin/activate
python -m pip install -e '.[analysis,figures,test]'
python -m pip install -e '../TileScope/python[test]'

The last command assumes the TileScope core checkout is beside this repository.

Where to put the data

Projects/
  TileScope/
  TileScope-results/               # code; generated artwork in analyses/figures/
  TileScope-data/
    input_data/
    supplementary_data/
      Supplementary_Data_1/
      Supplementary_Data_2/
      Supplementary_Data_3/
      Supplementary_Data_4/

Set TILESCOPE_DATA_DIR or TILESCOPE_CORE_DIR for different locations. No machine-specific absolute paths are required. Downloaded data never need to be placed inside TileScope-results.

python -m tilescope_results.bundles all
python -m analyses.figures.build all
# Or one group: overview, benchmark, greenwald, mouse, xenium

Builders read the bundles without changing them. Individual SVG/PDF panels, PNG previews and figure-specific CSV/Excel source data go to analyses/figures/. Set TILESCOPE_FIGURE_DIR to use a separate output directory.

Tests and documentation

python -m pytest tests/scenario_benchmark
python -m pytest analyses/figures -q
python -m pytest tests/e2e --run-real-data

Figure tests need the downloaded bundles. The optional integration test uses the small 242_T fixture.

• Data access: scope, source studies and deposit status.
• Data layout: paths and reproduction commands.
• Figure generation: individual panels and source data.
• Scenario benchmark and contract.