README.md
In ncchung/SeuratAddon: Unsupervised evaluation of cell identities in Seurat

SeuratAddon: Unsupervised evaluation of clustering-based cell identities

This R package helps the user to apply and visualize unsupervised evaluation of cell identities using a Seurat object, using the jackstraw approach. Cell identities of single cell samples derived from heterogeneous populations are often determined by clustering scRNA-seq data. Clustering-based cell identities are used in downstream analysis, feature selection, and visualization. The jackstraw methods enable us to examine if individual cell identities are accurately inferred using clustering.

SeuratAddon adds new functionalities that can be used with a Seurat object. K-means clustering is enhanced with the Kmeans++ initialization for faster convergence and the mini-batch algorithm that can easily scale to millions of single cells. Cell identities that are determined by clustering algorithms are evaluated by the jackstraw methods, providing p-values and posterior inclusion probabilities (PIPs) for individual single cells. These probabilities can be used for feature selection and visualization.

The tutorial (R package vignette) is available as a HTML or PDF document.

Installation

To use a development version from GitHub:

install.packages("devtools")
library("devtools")
install_github("ncchung/SeuratAddon")

Resources

Make sure you have installed Seurat and its dependencies. There are good tutorials for Getting Started with Seurat

The key feature of SeuratAddon is the capability to run the jackstraw method for cluster assignments on single cell RNA-seq data. For more information about the jackstraw, please see the jackstraw R package