Fig. 4: scGALA Unifies Multi-omics Datasets Through Enhanced Cell Alignment Beyond Conventional Matches. | Nature Communications

Fig. 4: scGALA Unifies Multi-omics Datasets Through Enhanced Cell Alignment Beyond Conventional Matches.

From: scGALA advances graph link prediction-based cell alignment for comprehensive data integration and harmonization

Fig. 4: scGALA Unifies Multi-omics Datasets Through Enhanced Cell Alignment Beyond Conventional Matches.

a Schematic workflow of scGALA integrating two dual-omics datasets (e.g., RNA+ATAC and RNA+ADT) into a unified tri-omics dataset via refined cell alignment. b Cell type-specific alignment precision: Confusion matrices comparing MNN (baseline), scGALA-enhanced (MNN + scGALA-exclusive), and scGALA-exclusive alignments. Diagonal enrichment highlights scGALA's improved accuracy in tri-omics integration. c Spearman’s correlation distribution of aligned cell pairs from MNN (baseline) versus scGALA-exclusive matches. Comparable distributions confirm scGALA maintains alignment quality while expanding coverage. d Composition of scGALA alignments (doughnut chart): scGALA identifies previously undiscovered cell pairs (exclusive to scGALA) while retaining high-confidence MNN matches, enabling comprehensive multi-omics unification. e ROC curves quantifying alignment accuracy for cell type matching, with AUROC values demonstrating scGALA's superior performance in cross-modality integration. f UMAP visualization of cross-dataset alignment, with connecting lines indicating alignments from MNN (baseline), scGALA-enhanced (MNN + scGALA-exclusive), and scGALA-exclusive alignments (blue: correct cell type matches; pink: mismatches). Source data are provided as a Source Data file.

Back to article page