REMI is designed to predict ecosystem-wide ligand receptor interactions within a microenvironment given RNA-sequencing data. The method moves beyond pairwise interactions and accounts for the effect of 2+ cell types on system-level interactions. Specifically, it creates communities of multicellular genes and identifies which interactions have a high probability of being transcriptionally linked.
Manuscript: https://www.science.org/doi/10.1126/sciadv.abi4757
Install R (>= 3.6)
install.packages("devtools")
library(devtools)
install_github("plevritis-lab/remi")
REMI takes in normalized bulk or single-cell RNA-sequencing data as an input, where the columns are samples and rows are genes. The column names are labeled as sample_celltype (i.e. S01_Bcell). The package has a built-in option to filter low-expressed genes, but it can also take it any pre-filtered scaled datasets (filter=F). For single-cell RNA-sequencing data, REMI can be run directly from the Seurat object.
vignettes
directory of the repo):To use a different ligand-receptor database (i.e. mouse), set the variable lr.database as the new LR data table. Column names must be Pair.Name, Ligand, and Receptor where Pair.Name separates the Ligand and Receptor by "_". Gene symbols in uploaded table must match gene names in input file.
new.lr.table <- read_csv("newlrtable.csv")
new.lr.table
Pair.Name Ligand Receptor
<chr> <chr> <chr>
1 A2M_LRP1 A2M LRP1
2 AANAT_MTNR1A AANAT MTNR1A
3 AANAT_MTNR1B AANAT MTNR1B
remi.res <- remi(obj, lr.database = new.lr.table)
Email: Alice Yu (ayu1@alumni.stanford.edu) or Sylvia Plevritis (sylvia.plevritis@stanford.edu)