Grinn

fetchGrinnCorrNetwork

This example shows how to reconstruct a biological network (grinn network) using information from Grinn internal database, then compute and combine with a weighted correlation network

INPUT

The table summarizes important arguments

Argument	Value	Description
txtInput	list of keywords e.g. ENSG00000143811	Keywords are IDs from a specific database.
from	type of keyword e.g. metabolite	Type of starting points in the network. It can be one of metabolite, protein, gene, pathway.
to	type of end nodes e.g. pathway	Type of endpoints in the network. It can be one of metabolite, protein, gene, pathway.
datX	data frame containing normalized-omic data	Columns correspond to entities e.g. genes and rows to samples e.g. normals, tumors. Require 'nodetype' at the first row to indicate the type of entities in each column.
datY	data frame containing normalized-omic data	It uses the same format as datX. Or it can be NULL, if there is only one omic dataset
corrCoef	numerical value from 0-1	The minimum absolute correlations to include edges in the output
pval	numerical value	The maximum pvalues, to include edges in the output

EXECUTE FUNCTION

Build a grinn network of pathway-metabolite using the list of SMPDB pathways, then combine with a correlation network of metabolites


#1. load metabolomics data from a csv file
datMet = read.csv("Lung_MET.csv", header=TRUE, row.names=1, stringsAsFactors=FALSE)
#2. show dimensions of metabolomics data
dim(datMet)
#3. show the first 10 rows and 10 columns
datMet[1:10,1:10]

#!!--- NOTE: The following codes convert kegg ids to grinn ids. If the input data is already the grinn ids, these steps can be skipped.
grinnID = convertToGrinnID(txtInput=colnames(datMet), nodetype="metabolite", dbXref="kegg") #call grinn function to convert ids
grinnID = grinnID[!duplicated(grinnID[,1]),] #keep the first mapped id
colnames(datMet) = lapply(colnames(datMet),function(x) ifelse(length(which(grinnID$FROM_kegg == x))>0,as.character(grinnID$GRINNID[which(grinnID$FROM_kegg == x)]),x))
#---------- END id conversion ----------#

#4. load pathway keywords from a text file (.txt)
ptwKw = unlist(read.csv("pathwayList.csv", stringsAsFactors=FALSE, header=FALSE))
#5. show length of keywords
length(ptwKw)
#6. show the first 2 keywords
ptwKw[1:2]
#7. execute function
result <- fetchGrinnCorrNetwork(txtInput=ptwKw, from="pathway", to="metabolite", datX=datMet, corrCoef=0.7, pval=1e-5, method="spearman", dbXref = "smpdb")
#display the first 10 edgelists
result$edges[1:10,]

EXPORT OUTPUT

Export the network as tab-delimited files to visualize in Cytoscape


write.table(as.matrix(result$edges),"grCorrNwEdge.txt",sep="\t",row.names = F, quote = FALSE)
write.table(as.matrix(result$nodes),"grCorrNwNode.txt",sep="\t",row.names = F, quote = FALSE)

VISUALIZATION
The figure is generated by Cytoscape 3.1.1 using grinn style (grinn.xml). It is corresponding to the cytoscape file grCorrNw.cys.

Diagram legend

References

Metabolomics and transcriptomics data used in this example are taken from the following publication:

Wikoff WR, et al. Metabolomic markers of altered nucleotide metabolism in early stage adenocarcinoma. Cancer Prev Res (Phila) 2015;8(5):410-8.