LoginSign Up
← Back to all sparks
G

glystats

ANALYTICS
Velocity0.0

Statistical Analysis of Glycoproteomics and Glycomics Data

github.com/glycoverse/glystats ↗

glystats keeps deleting analyses it decided belong somewhere else.

glycomicsstatisticsdifferential analysisbreaking changesbioconductor
◆Current state
glystats provides the statistical tests behind glycoverse: differential analysis, clustering, dimensionality reduction, and until recently enrichment. The last four releases are mostly subtraction. WGCNA and consensus clustering were removed in 0.10.0 as too interactive for a pipeline package, the enrichment functions were deprecated in the same release and deleted in 0.11.0 in favor of glyfun, and 0.11.0 also removed every gly_*_() matrix and vector interface. What remains accepts SummarizedExperiment inputs.
◆Where it's heading
The package is narrowing on purpose. The through-line across removals is a refusal to support two calling conventions or two homes for the same analysis: containers only, no bare matrices; one enrichment implementation, in glyfun. The additive work in the window went to statistical rigor rather than surface area, with effect sizes added to the four main tests, sign bugs fixed, and the log2 pseudo-count reduced.
◆Prediction
With the interface pruning finished, the next releases are more likely to deepen the tests that remain than to add new analysis families.

◆Recent moves

  1. 1mo ago

    Docs recommend the new SE containers

    Documentation and vignettes switch to GlycomicSE and GlycoproteomicSE with SummarizedExperiment accessors, closing glystats's part of the Stage II migration.

    View source ↗
  2. 1mo ago

    Tests recognize both container types consistently

    The statistical functions detect legacy experiment() and SummarizedExperiment inputs consistently across supported glyexp versions, so results do not depend on which container a user happens to be holding mid-migration.

    View source ↗
  3. 1mo ago

    Matrix interfaces and enrichment functions removed

    The largest cut in the window: every gly_*_() matrix and vector interface is removed, and the seven enrichment functions deprecated in 0.10.0 are deleted in favor of glyfun. In exchange, all remaining gly_*() functions and filter_sig_vars() accept SummarizedExperiment objects.

    View source ↗
  4. 2mo ago

    Smaller log2 pseudo-count across DEA and clustering

    A smaller pseudo-count reduces bias for low-intensity values in differential and clustering functions. A numerical adjustment rather than a behavioral change, though it will shift borderline results.

    View source ↗
  5. 3mo ago

    WGCNA and clustering dropped, enrichment deprecated

    gly_wgcna() and gly_consensus_clustering() are removed with users pointed at the underlying packages directly, and all enrichment functions are deprecated toward glyfun. This is the release where the scope decision is made; 0.11.0 only executes it.

    View source ↗
  6. 4mo ago

    Effect sizes added and test statistic signs fixed

    Cohen's d, rank-biserial correlation, eta-squared, and epsilon-squared join the tidy results for the four main tests, get_tidy_result() becomes type-stable, and incorrect statistic signs in gly_ttest() and gly_wilcox() are fixed. The sign bug is the kind that silently inverts published conclusions.

    View source ↗