Skip to contents

cellspecR

cellspecR defines the cellspec 1.0 table contract for segmented cells from multiplexed tissue images. It reads tool exports, validates structure and semantics, and writes an integrity checked directory for downstream analysis.

Status: 1.1.0. The changes since 1.0.0 are listed in NEWS.md. The cellspec 1.0 specification and the package API are stable. Readers that cs_formats() marks as experimental may still change with the export formats of their tools. cellspecR is not yet on CRAN.

Installation 

# install.packages("pak")
pak::pak("CTTIR/cellspecR")

Quick start 

x <- cs_example()
print(x)
#> <cellspec> spec 1.0.0
#>   cells     80 in 2 images (2 samples)
#>   features  19: 16 intensity, 2 shape, 1 other
#>   markers   4: DAPI, CD3e, Pan-Cytokeratin, FOXP3
#>   pixel     0.5 um/px
#>   adjacency none
#>   source    simulate 1.0.0 (cellspecR simulator 1.0.0)
cs_validate(x)
#> <cellspec validation> 45 checks: 0 fail, 0 warn, 0 skip, 45 pass
#> pass 45 other checks

policy <- cs_signal_policy(
  marker = c("CD3e", "FOXP3"),
  compartment = c("cell", "nucleus"),
  statistic = "mean"
)
signal <- cs_signal_matrix(x, policy)
dim(signal$signal)
#> [1] 80  2

destination <- file.path(tempdir(), "cellspec-example")
cs_write(x, destination, overwrite = TRUE)
cs_verify(destination)
#>                        file
#> cells.parquet cells.parquet
#> cellspec.json cellspec.json
#> 1                      DONE
#>                                                                       expected
#> cells.parquet 67ca57b984c08b81087d62183e3602dac68637447b40f3a7bf0697bc98557994
#> cellspec.json 990e5555e66725fefc2bd165df7022ed3aa217ed2a47df2b57e654a33986cdb9
#> 1             a6fe4004b01506e040b1f276afdac06d70416451a31a9fbf598536be3509f9de
#>                                                                       observed
#> cells.parquet 67ca57b984c08b81087d62183e3602dac68637447b40f3a7bf0697bc98557994
#> cellspec.json 990e5555e66725fefc2bd165df7022ed3aa217ed2a47df2b57e654a33986cdb9
#> 1             a6fe4004b01506e040b1f276afdac06d70416451a31a9fbf598536be3509f9de
#>                 ok
#> cells.parquet TRUE
#> cellspec.json TRUE
#> 1             TRUE
y <- cs_read_cellspec(destination)
identical(cs_cells(x), cs_cells(y))
#> [1] TRUE

What the package provides

  • cs_read() and cs_detect_format() for tool exports, plus cs_column_map() for explicit generic tables.
  • cs_new(), accessors, subsetting and compact print and summary methods.
  • cs_validate() and cs_feature_support() for structural and signal quality checks.
  • Parquet or exact text storage with a JSON sidecar, SHA-256 manifest and atomic staging through cs_write() and cs_read_cellspec().
  • Explicit signal policies, marker maps, object binding and tiled ownership.
  • Optional conversion to SpatialExperiment and AnnData, quality-control plots and a local Shiny review app.

The package does not segment images, normalize signals, gate cells or perform spatial statistics. It keeps the table contract and provenance at the boundary between those tools.

Data and reproducibility

The bundled export in inst/extdata/ is synthetic and contains no patient data. Reader fixtures record their generator, header grammar and SHA-256 in tests/testthat/fixtures/README.md. Examples do not access the network.

Contributing and license

Contributions follow the CTTIR contributing guide. The package is released under the MIT license.

Use of LLM tools

Where applicable, LLM tools were used for narrowly defined, non-authorial tasks: copyediting, prose refinement, Markdown/LaTeX formatting, boilerplate scaffolding (such as CI configurations and build scripts), and code refactoring. Tools included Chat AI, the KISSKI (GWDG) service, and self-hosted Mistral Small (24B, Apache-2.0) served through a local vLLM architecture on an Ubuntu Linux workstation with Intel processors and NVIDIA GPUs. The self-hosted model ran locally, with no data sent to third parties.

All scientific claims, methodological choices, analyses, interpretations and conclusions remain the authors’ own. Where applicable, LLM-generated material must be reviewed and revised before inclusion, and references must be verified against their DOI, arXiv ID or ISBN.

Local software setup: vLLM with Python/PyTorch, plus R and RStudio.

Branding

Editable R sources and regeneration instructions are in tools/branding.