NEWS

phynotype 0.6.2 (2026-09-01)

plot_biplot() no longer depends on factoextra. It's now a self-contained reimplementation of fviz_pca_biplot()/fviz_mca_biplot()'s geometry (individual points/arrows, variable arrows or category points, axis percentage labels, dashed origin lines, the r * 0.7 variable-arrow scaling rule, and the exact cos2/contribution formulas factoextra uses for prcomp/FactoMineR::MCA objects), verified to reproduce numerically identical coordinates. factoextra dropped from Suggests; ggrepel added to Imports (previously an indirect factoextra dependency, now used directly for repelled labels).

Standardized figure colors package-wide. theme_phynotype() now builds on ggplot2::theme_classic() with an Okabe-Ito colorblind-safe palette (matching the funcml package's theme_funcml()), and new scale_color_phynotype()/scale_fill_phynotype() discrete scales are applied to every plot with a categorical color/fill mapping (cluster embedding, consensus clusters, feature profiles, silhouette widths). Previously these fell back to ggplot2's default hue palette.
plot(ceteris_paribus) no longer uses its own ad hoc teal/purple colors (visually indistinguishable from the DALEX/ingredients package's ceteris-paribus plot style, despite no dependency on it) - it now uses the same blue/vermillion accent pair as the rest of the package.
plot_coassoc()'s co-association heatmap now uses a proper continuous viridis gradient instead of ggplot2's default blue gradient.
plot(lime_explanation)'s positive/negative bars now use a consistent green/vermillion pair instead of ggplot2's default hue palette.
The two independently hardcoded feature-importance/cluster-size bar colors are now sourced from the same shared accent constant.
README figures re-rendered to reflect the new palette.

Added plot_biplot(), a thin wrapper around factoextra::fviz_pca_biplot()/fviz_mca_biplot() for "pca" and "mca" embeddings. No FAMD or MDS equivalent, matching what factoextra itself provides. Three variants, all following factoextra's own documented examples: "cluster" (default, colored/shaped by cluster assignment), "cos2" (colored by quality of representation), and "label" (individuals shown as text instead of points).
Added theme_phynotype(), a single ggplot2 theme now used consistently across every plot function in the package (previously some used theme_minimal(), one used theme_bw(), and plot(ceteris_paribus) had its own ad hoc purple-tinted theme).
Fixed metacluster() so it can combine native mixed-data methods (kproto, kmm) with numeric-only (kmeans, pam, dbscan, gmm) or distance-based (hclust, agnes) methods on the same mixed-type dataset, by preparing per-method input instead of passing raw mixed data to every candidate.
Fixed validate.metacluster_fit() to encode mixed-type data before computing silhouette/Calinski-Harabasz/Davies-Bouldin, matching validate.cluster_fit().
All numeric package results now round to 4 significant digits by default: validate()'s metrics/per-cluster tables, explore()'s feature summary/separation/prototype tables, centers()/prototypes(), feature_importance(), ceteris_paribus(), lime_explain(), and metacluster()'s selection/stability summaries.
First CRAN submission.

Validation tables now include metric scale and preferred direction metadata for silhouette, bootstrap ARI, Davies-Bouldin, total within-cluster sum of squares, and Calinski-Harabasz.
plot_validation() now facets metrics into individual panels instead of combining incompatible scales on one axis.
Added the initial cluster() workflow with kmeans, pam, hclust, agnes, dbscan, and gmm wrappers.
Added metacluster() with co-association consensus clustering.
Added validate(), explore(), predict(), and core plotting helpers.
Added package tests, documentation, and starter vignettes.