BICCN Marmoset

Report card for Correlation, Tree, and Seurat Mapping on Marmoset motor cortex (Zemke et al. 2023)

Overview

A taxonomy was initially built using the BICCN’s marmoset motor cortex single nucleus 10x Multiome dataset. In building the taxonomy, 1000 binary marker genes were selected based on their gene expression from the single-cell transcriptome. Subsequently, the dataset was mapped to itself, termed self-projection, for evaluating the ideal performances of correlation, tree, and seurat mapping algorithms.

Quantitative analysis

The analysis evaluates the predictions of correlation, tree, and Seurat mappings in determining cluster labels in a self-projection evaluation.

Annotaion	F1-score
Cluster Correlation Mapping	0.761
Cluster Tree Mapping	0.711
Cluster Seurat Mapping	0.972

Correlation Mapping

1. Label-wise F1-score
   

2. Confidence values for correctly and incorrectly assigned labels
   

3. Confusion matrix (row-normalized)
   

Tree Mapping

1. Label-wise F1-score
   

2. Confidence values for correctly and incorrectly assigned labels
   

3. Confusion matrix (row-normalized)
   

Seurat mapping

1. Label-wise F1-score
   

2. Confidence values for correctly and incorrectly assigned labels
   

3. Confusion matrix (row-normalized)
   

</details>