NEWS

CellMentor 1.1.2 (2026-05-22)

calculate_performance() now passes as.factor(true_labels) and as.factor(clusters) to aricode::NMI(). As of aricode 1.1.0 (currently shipping in Bioconductor 3.24), NMI() routes mixed-type inputs through sort_pairs(), which calls as.integer() on character labels — turning cell-type names such as "alpha" into NA and aborting with "NA are not supported.". Coercing both sides to factor keeps the computation well-defined and restores compatibility with the Bioconductor build farm. This was the underlying cause of the vignette failure that 1.1.1 surfaced but did not fix.

cluster_nmf() now falls back to k-means clustering when the NMF embedding has too few features or cells for SNN (variable-feature selection on low-dim embeddings was failing under Seurat 5.x). The number of k-means / hclust centers is now clamped to a feasible range, so grid search no longer fails silently on small reference splits.
CellMentor() now stops with a clear message if every parameter configuration in the grid search fails, instead of cascading to a cryptic switch() error during final-model training.
evaluate_params() warnings now include the underlying error message, so build logs surface the real cause of failures.
initialize_wh() validates that method is a single character string and reports a helpful error if not.

Updated citation to the published Nature Communications paper (Hevdeli et al., 2026, doi:10.1038/s41467-025-67088-7).
Added inst/CITATION so citation("CellMentor") returns a structured bibliographic entry.