Solutions - Bayesian Networks in R: With Applications in Systems Biology - page 136

> sum(DBNlasso != 0)/prod(dim(DBNlasso))
[1] 0.1753436
> plot(sort(abs(DBNlasso), decr = TRUE)[1:500],
+ type = "l",
+ ylab = "Absolute coefficients")

• James-Stein shrinkage with GeneNet :

> DBNGeneNet = ggm.estimate.pcor(dataVar2,
+ method = "dynamic")
> # p-values, q-values and posterior probabilities
> # for each potential arc
> DBNGeneNet.edges =
+ network.test.edges(DBNGeneNet)
> # plot the arcs probability by decreasing order
> plot(DBNGeneNet.edges[, "prob"], type = "l")
> # number of arcs with prob > 0.9
> # (i.e. local fdr < 0.1)
> sum(DBNGeneNet.edges$prob > 0.9)

• First-order conditional dependencies approximation with G1DBN :

> library(G1DBN)
> data = dataVar2inline
> # next step is a bit long but is less than
> # 3 minutes with a regular PC
> DBNG1DBNstep1 =
+ DBNScoreStep1(data, method = "ls")
> DBNG1DBN = DBNScoreStep2(DBNG1DBNstep1$S1ls,
+ data, method = "ls", alpha1 = 0.5)
> plot(sort(DBNG1DBN), type = "l",
+ ylab = "Arcs' p-values")

• LASSO and network modularity with SIMoNe :

> library(simone)
> data = dataVar2inline
> ctrl = setOptions(clusters.crit = "BIC")
> DBNsimone.BIC =
+ simone(data, type = "time-course",
+ clustering = TRUE, control = ctrl)
> DBNsimone.BIC.net = getNetwork(DBNsimone.BIC)
> # number of arcs:
> sum(DBNsimone.BIC.net$A == 1)

3.3 Consider the dimension reduction approaches used in the previous exercise
and the arth800 data set from the GeneNet package.