Hi,

I have used bioconductor package FEM on my 450k methylation and RNAseq data. I followed the instruction described in the vignette. I can manage to run DoEpiMod and DoExpMod functions on the methylation and RNAseq datasets successfully independently. However, when I tried to run DoFEMbi on the same datasets, I got the following error: 

It seems to process through the "Constructing weighted network" step successfully but fails at the "Running Spin-Glass algorithm" step. Any suggestions to overcome this error ? Thanks a lot.

[1] "Running Spin-Glass algorithm"

Error in .Call("R_igraph_spinglass_my_community", graph, weights, as.igraph.vs(graph,  : 
  At clustertool.cpp:472 : Cannot work with unconnected graph, Invalid value

Here is the code that I used:

intFEM.allProbe <- list(statM=statM.o.mval.dm,
                        statR=statR.o.voom.de,
                        adj=hprdAsigH.m)

fembi.allProbe <- DoFEMbi(g1FEM.allProbe,
                   nseeds=100,gamma=0.5,nMC=1000,sizeR.v=c(1,100),
                   minsizeOUT=10,writeOUT=TRUE,nameSTUDY="g1-allProbe_fem",ew.v=NULL);

Hi Sirintra,

Your question is easily answered as follows:

The problem you are encountering relates to the fact that the spin-glass algorithm can't be run on an unconnected network. The network that you are passing on
to the function DoFEMbi, specified by the adjacency matrix hprdAsigH.m, has not been correctly integrated with the DNA methylation and gene expression data, so could result in an unconnected network.

You need to do the proper integration first, before running DoFEMbi function. The correct flow is as follows:

statM.o <- GenStatM();
statR.o <- GenStatR();
int.o <- DoIntegFEM450k(statM.o,statR.o,hprdAsigH.m,cM=1,cR=1);
fem.o <- DoFEMbi(int.o,.....)

rgds

Andrew T,