-
Evaluation CSV: Long-format table with columns: tool, cell_type, pearson, spearman, concordance, rmse, mad, percentdeviation
-
Predictions CSV: Wide-format table with columns: sample, tool, cell_type_1, cell_type_2, ..., cell_type_n
-
Ground-Truth file: Wide-format table with columns: sample, cell_type_1, cell_type_2, ..., cell_type_n
-
bland_altman: Shows mean bias and ±1.96 SD limits to visualise agreement spread for assessing systematic and proportional error -
error_plot: Shows prediction‑minus‑truth bars per sample for identifying outlier samples driving total error -
stacked_bar_proportions: Shows predicted cell‑type stacks per sample for identifying unrealistic totals or proportions -
regression_plot: Shows actual vs predicted with identity and fit line for checking slope and intercept bias -
response_plot: Shows truth and prediction dots across samples for testing preservation of relative abundance patterns
-
annotated_heatmap: Shows correlation and deviation heatmaps side by side for handy precision and accuracy report per cell type -
boxplot_proportions: Shows distribution of predicted proportions per cell type for identifying spread, skew, and outliers -
cell_type_analysis: Shows combined error, correlation, error ranking, and tool heatmap for quick summary of cell‑type‑specific difficulties -
metrics_by_tool_heatmap: Shows metric values per-cell for highlighting tool‑specific performance gaps -
metrics_comparison: Shows boxplots of metric distributions across cells for comparing median performance and variability -
metrics_correlation: Shows inter‑metric correlation matrix for identifying redundant or inversely related metrics -
predicted_vs_actual: Shows scatter grids of actual vs predicted for checking overall calibration consistency -
true_stacked_bar_proportions: Shows ground-truth proportions for comparing tostacked_bar_proportions