In shanlu01/mlg: MLG Clustering

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/",
  out.width = "100%"
)

What is MLG?

MLG is an integrative clustering approach for single cell RNA-seq data across multiple experimental conditions. MLG takes multiple low-dimensional embedding data of gene expression matrix as input, e.g. low dimension embeddings from dimension reduction methods like PCA, cNMF, or from data integration methods like Seurat, Liger. It constructs a multi-layer shared nearest neighbor (SNN) graph from these low-dimensional embeddings and performs Louvain graph partitioning algorithm.

knitr::include_graphics("man/figures/mlg_workflow.png")

Installation tips

The vignettes of mlg package depend on R package Liger (Linked Inference of Genomic Experimental Relationships) and Seurat. After installing dependencies, you can install mlg with the following commend:

devtools::install_github('shanlu01/mlg')

What can you do with mlg?

1. Perform MLG clustering

mlg_cluster(
  factor.list,
  cluster.resolution
  )

1. Visualize the MLG graph through force directed layout.

mlg_visualization(
  factor.list,
  label
)

Argument factor.list is a list variable containing low dimensional embedding data, for example let factor.list = list(PCA_factors, cNMF_factors). cluster.resolution is the resolution in modularity maximization. A larger resolution number will lead to more clusters. Argument label in mlg_visulization is the color labeling to be imposed on the figure. For example, to visualize clustering result, we can specify label = clusters.

Contact

The package is developed in Keles Research Group at University of Wisconsin - Madison. Please contact Shan Lu ([e-mail]) or open an issue in the github repository for any question and suggestion.

shanlu01/mlg documentation built on Oct. 20, 2020, 6:36 a.m.