Fig. 3: Performance of GRNBoost2 in recovering the imposed edges versus unimposed edges (PBMC-CTL). | Nature Communications

Fig. 3: Performance of GRNBoost2 in recovering the imposed edges versus unimposed edges (PBMC-CTL).

From: GRouNdGAN: GRN-guided simulation of single-cell RNA-seq data using causal generative adversarial networks

Fig. 3: Performance of GRNBoost2 in recovering the imposed edges versus unimposed edges (PBMC-CTL).

Left column shows the AUPRC (A) and Precision at k (per gene) (C) when the imposed edges (positive control GRN) were considered the ground truth. Right column shows the AUPRC (B) and Precision at k (per gene) (D) when the unimposed edges (negative control GRN) were considered the ground truth. Precision at k (per gene) refers to the precision when top k TFs for each gene is used to form the reconstructed GRN. All three datasets had the same number of cells (n = 19,773). Source data are provided as a Source Data file.

Back to article page