Variable Gene Selection Method

Use this option to select the method to calculate a subset of features that exhibit high cell-to-cell variation in the dataset. It has been reported that focusing on these genes in downstream analysis helps to highlight biological signal in single-cell datasets.

.Two methods are available:

Option:	Description:
Dispersion	Dispersion method is based on the log(variance/mean) plot in the Seurat R package1. It uses a relatively simple method to calculate gene variation while controlling for the strong relationship between variability and average expression. The dispersion method doesn't select genes based on the ranking of the dispersion alone; instead, it is based on both the dispersion & mean values. In other word, just because a gene is ranked high, based on dispersion alone, does not mean it is a good variable gene.
VST	VST (Variance-Stabilizing Transformation) uses Loess regression as a normalization method to control for the mean-variance relationship. It is also included in the the Seurat R package.

Option:

Description:

Dispersion

Dispersion method is based on the log(variance/mean) plot in the Seurat R package1. It uses a relatively simple method to calculate gene variation while controlling for the strong relationship between variability and average expression.

The dispersion method doesn't select genes based on the ranking of the dispersion alone; instead, it is based on both the dispersion & mean values. In other word, just because a gene is ranked high, based on dispersion alone, does not mean it is a good variable gene.

VST

VST (Variance-Stabilizing Transformation) uses Loess regression as a normalization method to control for the mean-variance relationship. It is also included in the the Seurat R package.

Please see Stuart, T. et al. 2019. Comprehensive Integration of Single-Cell Data. Cell 177:1888-1902. for more information.

To Specify the Gene Selection Method:

Select the method using the drop-down menu.