LoginSign Up
← Back to all sparks
S

scTypeEval

ANALYTICS
Velocity0.0

R package scTypeEval by carmonalab — release notes from GitHub.

github.com/carmonalab/scTypeEval ↗

scTypeEval judges single-cell annotations without needing a ground truth to judge them against.

single-cellcell type annotationbioconductorclustering evaluationr package
◆Current state
scTypeEval evaluates the consistency of single-cell cell type annotations without a ground-truth reference, using pseudobulk distances, Wasserstein distances and reciprocal classification as alternative dissimilarity strategies. It cleared Bioconductor's submission process on its first cycle, moving from a 0.99.30 pre-release in April 2026 to version 1.0.0 in the Bioconductor 3.23 release a month later. It accepts matrix, Seurat and SingleCellExperiment inputs.
◆Where it's heading
The visible history is short and shaped entirely by the Bioconductor pipeline — the 0.99.x series is that project's submission convention, and 1.0.0 is what acceptance looks like rather than a maturity claim by the authors. What the pre-release notes emphasise is breadth of input format and of dissimilarity strategy rather than a single recommended method, which suggests the package is positioned as a comparison harness rather than a scoring tool. Nothing in these two entries indicates work beyond getting accepted.
◆Prediction
With only a submission cycle in the record, there is not enough here to predict a direction; the next release will most likely be whatever the Bioconductor 3.24 cycle requires, and the first post-acceptance release is what will show whether development continues.

◆Recent moves

  1. 3mo ago

    Accepted into the Bioconductor 3.23 release

    The package's version-1.0.0 tag for the Bioconductor 3.23 release, marking acceptance out of the 0.99.x submission series. The entry is a single line and describes no code change.

    View source ↗
  2. 4mo ago

    Bioconductor pre-release: ground-truth-agnostic annotation evaluation

    The Bioconductor pre-release describing the package's scope: multiple dissimilarity strategies including pseudobulk distances, Wasserstein and reciprocal classification, internal validation metrics for annotation consistency across samples, and PCA, MDS and heatmap visualisation. Runnable vignettes cover matrix, Seurat and SingleCellExperiment workflows, alongside a snake_case API standardisation.

    View source ↗