Skip to main content


Springer Nature is making SARS-CoV-2 and COVID-19 research free. View research | View latest news | Sign up for updates

Fig. 2 | Genome Biology

Fig. 2

From: DENDRO: genetic heterogeneity profiling and subclone detection by single-cell RNA sequencing

Fig. 2

DENDRO accuracy assessment. a The overall simulation analysis pipeline. Mutation matrix (cell-by-loci) is generated according to a simulated evolutionary tree, where the leaves are subclones and mutations can be placed on the branches. Matrices of alternative allele (Xcg) and total read counts (Ncg) are sampled from a scRNA-seq dataset with known transcriptomic allele-specific read counts. DENDRO cluster is further applied, and its performance is assessed by adjusted Rand index (global accuracy), capture rate (subclone-specific sensitivity), and purity (subclone-specific precision). See Additional file 2: Supplementary Materials for detailed definition. Gray dashed line indicates optional input for DENDROplan, where bulk DNA-seq and bulk RNA-seq can guide the tree simulation and read count sampling procedure. b Cluster accuracy via simulation studies. Various parameters show effects on cluster accuracy (measured by adjusted Rand index) based on tree structure on the most right. Left panel: effect of mutation burden on fixed read depth. Right panel: effect of read depth on fixed mutation burden. c Evaluation of DENDRO on a renal cell carcinoma and its metastasis. (Left to right) (1) DENDRO clustering result from primary and metastatic renal cell carcinoma dataset. Background colors represent DEDRO clustering result. (2) Clustering of the same dataset using Z matrix (indicator matrix, Zij = 1 when detected a mutation for cell i at locus j by GATK tool). (3) Clustering of the same dataset using \( \frac{X}{N} \) matrix (mutation allele frequency matrix). (4) Clustering of the same dataset using expression (log(TPM + 1))

Back to article page