In ramellose/CoNetinR: CoNet implementation in R

# Global options
library(knitr)
opts_chunk$set(fig.path="figure_seqtime_tour/")

We start by loading the CoNetinR, vegan, seqtime and gdata libraries.

library(CoNetinR)
library(vegan)
library(gdata)
library(seqtime)

The next step is to generate an interaction matrix, and from it, a dataset.

N = 50
S = 40
A = generateA(N, "klemm", pep=10, c =0.05)
dataset = generateDataSet(S, A)

Next, we can use the CoNet adaptation to try and find back the original interaction matrix. We are going to use the Spearman method, and first get the Spearman scores.

scores = getNetwork(mat = A, method="spearman", T.up=0.2, T.down=-0.2, shuffle.samples=F, norm=TRUE, rarefy=0, stand.rows=F, pval.cor=F, permut=F, renorm=F, permutandboot=T, iters=100, bh=T, min.occ=0, keep.filtered=F, plot=F, report.full=T, verbose=F)
scores = scores$scores

We also need to get the p-values.

pmatrix = getNetwork(mat = dataset, method="spearman", T.up=0.2, T.down=-0.2, shuffle.samples=F, norm=TRUE, rarefy=0, stand.rows=F, pval.cor=T, permut=F, renorm=F, permutandboot=F, iters=100, bh=T, min.occ=0, keep.filtered=F, plot=F, report.full=T, verbose=F)
pmatrix = pmatrix$pvalues

Of course, now we have the Spearman correlations and the p-values. We can turn that into an adjacency matrix.

adjmatrix = matrix(nrow = N, ncol = N)
adjmatrix[lower.tri(adjmatrix)] = scores
adjmatrix = t(adjmatrix)
adjmatrix[lower.tri(adjmatrix)] = scores
for (i in 1:N){
   for (j in 1:N){
    if (is.na(adjmatrix[i,j])){
       adjmatrix[i,j] = 0
    }
    else if (pmatrix[i,j] > 0.05){
       adjmatrix[i,j] = 0
    }
   }
 }