Fig. 3 |. Simulation studies for comparing BayesCNet, which leverages cell type relationship inferred from cell ontology tree (COT), to competing methods for inferring enhancer-gene linkages.

a. A simulated cell ontology tree, mimicking PBMC cell type hierarchy, for simulation. Two target cell types were selected to represent different connectivity levels: Cell type 2 (low connectivity) and Cell type 6 (high connectivity). Connectivity scores are indicated at each node. b. Experimental design of the 2×3 factorial simulation for connectivity and sample size. Each target cell type was simulated at three sample size levels: low (5), medium (20), and high (60) metacells, and two connectivity level: low and high. The sample sizes of other cell types are fixed at the medium level. c. AUPRC across 100 simulations per method per scenario. d. AUROC across 100 simulations per method per scenario.