IRRG:a method for constructing cellular communication networks by integrating receptor-regulated gene expression information
by Shuqi Guo (guoshuqi9805@gmail.com),Shaowu Zhang* (zhangsw@nwpu.edu.cn),Yan Li (linay0124@outlook.com), Shihua Zhang (zsh@amss.ac.cn)
In this repository, We provide an R package for integrating receptor-regulated gene expression information on the basis of ligand-receptor interactions to construct cellular communication networks between cell typese.
├─Data [The database used by IRRG and the single cell data from the case study]
│ │ LRdb.rda [Ligand-Receptor Database]
│ │ mm2Hs.rda [List of mouse and human homologous gene interconversions]
│ │ PwC_ReactomeKEGG.rda [Signaling Pathway Database]
│ │
│ ├─Mouse_IFE [Mouse interfollicular epidermal scRNA-seq dataset]
│ │ cell_ann.csv [Cell type annotations of IFE]
│ │ IFE_express.csv [Gene expression matrix of IFE]
│ │
│ └─RCC [Renal cell carcinoma scRNA-seq dataset]
│ SI_22368_tumor.csv [Gene expression matrix of tumor tissues]
│ SI_22369_normal.csv [Gene expression matrix of normal tissues]
│
└─Program [The code of IRRG framework]
cal_Rscore.cpp [The function uses a random walk algorithm to iteratively update the receptor gene regulatory network]
data_process.R [Script for simple pre-processing of gene expression matrix]
IFE_Demo.R [Example of building a cellular communication network for IFE]
intra_network.R [The function used to build the receptor gene regulatory network]
LRscore.R [The function used to calculate the ligand-receptor interaction score]
net2adj.R [This function converts the network into an adjacency matrix]
R_info.R [The function used to calculate the receptor score]
separate_data.R [This function groups gene expression matrices according to whether the receptor is expressed or not]
simplify_interactions.R [This function simplifies the types of gene interactions in signaling pathways]
We provide an example of IRRG applied to the IFE dataset Program/IFE_Demo.R, which consists of the following steps.
1.Data pre-processing
Here we first use a simple R scriptProgram/data_process.R to normalize the gene expression matrix for pre-processing, the rows of the expression matrix represent genes and the columns represent cells.
2.Calculation of receptor scores
The receptor scores in each cell type can be calculated using the following function，and the results will be saved in the "R_info" folder under the working path.
R_info(data = data,cluster = cluster,c.names = c.names,cell.prop = 0.2,
LRdb=LRdb,mm2Hs = mm2Hs,PWC = PwC_ReactomeKEGG,species = "mus musculus")This function consists of two main steps: 1) call function Program/intra_network.R to build a gene regulatory network for each receptor in each cell type; 2) use function [Program/cal_Rscore.cpp] to calculate the stability value of the receptor nodes in the network in order to normalize for the receptor score.
3.Calculation of ligand-receptor interaction scores
The ligand-receptor co-expression was multiplied by the receptor score using the following functionProgram/LRscore.R to obtain all ligand-receptor interaction scores between each cell type.
LRscore(data = data,c.names = c.names,species = "mus musculus",celltype = celltype,Permutation.test = TRUE)The results will be saved in the "LR_score_notest" folder in the current path.
By setting the parameter "Permutation.test = TRUE", we can screen for significantly specific ligand-receptor pairs between cell types via the permutation test,and the results will be saved in the "LR_score_test" folder of the current path.
4.Cellular communication network construction
After the calculation in step 3, we will find a summary of the intercellular communication strength and the number of ligand-receptor pairs in the "Celltype_communication_summary" folder of the current pathway