search

carter-allen / banyan

Bayesian network models for community detection in spatially resolved single cell data
1 stars 0 forks source link

readme

banyan

Bayesian network models for spatially-resolved single-cell data

The banyan package allows for network-based Bayesian analysis of spatially resolved single cell data. The core model is an extension of the stochastic block model (SBM) to integration of spatial and gene expression cell-cell similarity network. The package allows for robust detection of cell sub-populations and interaction among them using a Bayesian probabilistic framework.

Installation

The banyan package can be installed directly from this repository using devtools.

devtools::install_github("carter-allen/banyan")

Requirements

The banyan package has multiple dependencies, one of which being mlsbm. The mlsbm package contains the core implementation of the Bayesian SBM written with Rcpp. You can install the development version of mlsbm from GitHub using:

# development version
devtools::install_github("carter-allen/mlsbm")

The igraph package is used to handle standard network routines and initialization. Note, due to the need for a resolution parameter in the initialization of cluster allocations with igraph::cluster_louvain() the development version of igraph must be installed. See here for more details.

We also allow for integration with standard Seurat workflows by accommodating Seurat objects as inputs.

Vignettes

Usage

Below are R commands for analyzing a publicly available anterior mouse brain data set published by 10X Genomics.

# packages
library(SeuratData)
library(Seurat)
library(banyan)
library(patchwork)

# load data
InstallData("stxBrain")
brain <- LoadData("stxBrain","anterior1")

# run Seurat processing steps
brain <- SCTransform(brain, assay = "Spatial", verbose = FALSE)
brain <- RunPCA(brain, assay = "SCT", verbose = FALSE)

# fit banyan model
brain_fit <- fit_banyan(seurat_obj = brain, K = 6)
save(brain_fit,file = paste0(data_dir,"brain_fit_banyan.RData"))
load(paste0(data_dir,"brain_fit_banyan.RData"))

# plot labels
plot_labels(brain_fit)

# calculate scores
brain_fit <- get_scores(brain_fit)

# plot uncertainty
plot_uncertainty(brain_fit)

# plot propensity
plot_propensity(brain_fit,k = 1) + plot_propensity(brain_fit,k = 2)

# plot connectivity matrix
plot_connectivity_matrix(brain_fit)

# plot connectivity interval
plot_connectivity_intervals(brain_fit)