DAG-MCTS for cell type annotation. Search space is a DAG (Merge + ReEmbed create multi-parent nodes). Uses P-UCT selection, Soft-Bellman backprop, confidence discounting, and transposition table. Per-iteration tree viz emitted as interactive HTML.
Correct batch effects in single-cell data using Harmony, BBKNN, or ComBat.
Query CellWiki knowledge base for marker genes, cell type ontologies, and tissue-specific cell type constraints.
Find differentially expressed genes (markers) for clusters using wilcoxon, t-test, or logistic regression. Includes consensus markers and resolution scanning.
Assess single-cell clustering quality using silhouette score, modularity, and phiclust. Returns recommendations to adjust resolution or accept current clustering.
Load single-cell RNA-seq data from h5ad, rds, or 10x mtx formats into AnnData objects.
Generate publication-quality UMAP/t-SNE embedding plots with Okabe-Ito color palette. Outputs PDF+PNG.
Generate violin plots and dot plots for gene expression visualization across cell clusters.