Estimate LUSC Mutation Networks

In this section, we estimate the relationships among mutated genes in 178 lung squamous cell carcinoma (LUSC) patients from the TCGA project. We obtained the data via getTCGA from TCGA2STAT package. A total of 13655 genes for LUSC patients were obtained. Genes with a mutation rate of less than 15% in the cohort or with an undefined gene name were filtered out before analysis. This left data with 59 genes and 179 patients. Similar to the work-flow applied on simulated data presented, Ising models were fit across 20 regularization values, and the optimal network was selected from 100 iterations via the StARS approach.

> lusc.mut <- getTCGA(disease="LUSC", data.type="Mutation")

> mut.dat <- lusc.mut$dat$gdats
> mut.rate <- apply(mut.dat, 1, sum)/ncol(mut.dat)
> mut.gd <- mut.dat[mut.rate>= 0.15, ]
> mut.gd <- mut.gd[-grep("Unknown", rownames(mut.gd)), ]

> lusc.mut.fit <- XMRF(mut.gd, method="ISM", N=100, nlams=20, 
                      stability="STAR", th=0.001, beta=0.1)
                      
> plotGML(lusc.mut.fit, fn="lusc.fit.gml", vars=rownames(mut.gd), weight=TRUE)

The estimated mutated gene networks for lung squamous cell carcinoma viewed from Cytoscape is shown in Figure S4.

Figure S4: LUSC mutated gene networks estimated by Ising model's XMRF(...,method="ISM")