Biology Reference
In-Depth Information
(a) For a comparative analysis of the different approaches, select the top
50
arcs for each approach (function
BuildEdges
from the G1DBN package
can be used to that end).
(b) Plot the four inferred networks with the function
plot
from package
G1DBN.
(c) How many arcs are common to the four inferred networks?
(d) Are the top
50
arcs of each inferred network similar? What can you con-
clude?
Using the results of the previous exercise.
(a) Top 50 arcs computation.
•
With the LASSO:
> lasso.thres.top50 = mean(sort(abs(DBNlasso),
+
decreasing = TRUE)[50:51])
> DBNlasso.50edges =
+
BuildEdges(score = -abs(DBNlasso),
+
threshold = -lasso.thres.top50)
•
With
GeneNet
:
> DBNGeneNet.50edges =
+
cbind(DBNGeneNet.edges[1:50, "node1"],
+
DBNGeneNet.edges[1:50, "node2"])
•
With
SIMoNe
:
> nbArcs = 50
> ctrl = setOptions(clusters.crit = nbArcs)
> DBNsimone.50 = simone(data,type = "time-course",
+ clustering = TRUE,control = ctrl)
> DBNsimone.50.net = getNetwork(DBNsimone.50,
+
selection = nbArcs)
> DBNsimone.edges =
+ BuildEdges(score = -DBNsimone.50.net$A,
+
threshold = 0)
•
With
G1DBN
:
> G1DBN.thres.top50 = mean(sort(DBNG1DBN)[50:51])
> DBNG1DBN.edges =
+ BuildEdges(score = DBNG1DBN,
+ threshold = G1DBN.thres.top50, prec = 3)
(b) Plots of the top 50 arcs identified by each approach.
•
With the LASSO:
> par(mfrow = c(2, 2))
> DBNlasso.top50 =
Search WWH ::
Custom Search