zhangxf-ccnu / panda

PANDA

R package supporting the paper “Dual decoding of cell types and gene expression in spatial transcriptomics with PANDA”.

PANDA is a probabilistic-based deconvolution method capable of simultaneously estimating cell type proportions and cell-type-specific gene expression for spatial transcriptomics data. Its unique feature lies in addressing variation within the same cell type during deconvolution, achieved through the integration of archetypal analysis. PANDA requires the gene expression count matrix along with cell type labels from the scRNA-seq reference, as well as the gene expression count matrix from the spatial transcriptomics data. The method involves two key steps. Firstly, PANDA employs archetypal analysis on scRNA-seq reference data to infer cell-type-specific archetypes, providing a comprehensive characterization of the state space for each cell type. Subsequently, PANDA performs deconvolution on spatial transcriptomics data to deliver accurate estimates of cell type proportions and cell-type-specific gene expression.

Installation

The PANDA package has the following R-package dependencies: Seurat, doParallel, foreach, parallel, umap, ggplot2.

The experiments implemented in the paper are based on Seurat v4.3.0, which can be installed as follows:

install.packages('remotes')
remotes::install_version(package = 'Seurat', version = package_version('4.3.0'))

All other R-package dependencies can be installed via install.packages.

After installing all dependencies, you can install the development version of PANDA as follows:

install.packages("devtools")
devtools::install_github("Zhangxf-ccnu/PANDA")

Usage

This is a basic example which shows you how to use PANDA for deconvolution.

Load package

library(PANDA)

Load datasets

• sc_counts: A matrix of the raw count expression in the scRNA-seq reference (cell x gene).
• sc_labels: A vector of the corresponding cell type labels in the scRNA-seq reference.
• st_counts: A matrix of the raw count expression in the spatial transcriptomics data (spot x gene).

Perform archetypal analysis on the scRNA-seq reference

sc_results <- sc_train(sc_counts, sc_labels, n_archetypes_vec = 10)

Perform deconvolution on the spatial transcriptomics data

st_results <- st_train(st_counts, sc_results = sc_results)

Extract results

The cell type proportions for spots can be extracted by

proportion <- st_results$proportion

The cell-type-specific gene expression can be extracted by

expression <- st_results$mu

Tutorial

A tutorial with examples of the usage of PANDA is available at: PANDA-examples.html.

Contact

Please do not hesitate to contact Mr. Meng-Guo Wang ([email protected]) or Dr. Xiao-Fei Zhang ([email protected]) to seek any clarifications regarding any contents or operation of the archive.