In RicUIB/AligNet: Tool to Analyze PPI Networks and Align Networks using AligNet Algorithm algorithm

How to use package PINTAS

library(knitr)

library(AligNet)

First we have to build the networks. We can build them using the function read.network. For example, from a list of protein interactions.

edges1 = matrix(c(
  "85962.HP0109", "85962.HP0136",
"85962.HP0109", "85962.HP0137",
"85962.HP0136", "85962.HP0247",
"85962.HP0136", "85962.HP0303",
"85962.HP0137", "85962.HP0247",
"85962.HP0137", "85962.HP0853",
"85962.HP0247", "85962.HP1316"
), ncol=2, byrow=TRUE)
hpy <- read.network(edges1, mode="edges")

plot(hpy)

edges2= matrix(c(
"DBP2_YEAST", "RL2A_YEAST",
"HAS1_YEAST", "MAK5_YEAST",
"NOP10_YEAST", "DBP2_YEAST",
"NOP10_YEAST", "HAS1_YEAST",
"NOP10_YEAST", "MAK5_YEAST",
"NOP10_YEAST", "RL2A_YEAST",
"TSA1_YEAST", "HSP7F_YEAST",
"TSA1_YEAST", "TSA2_YEAST"
), ncol=2, byrow=TRUE)
sce <- read.network(edges2,mode="edges")
plot(sce)

Now we need a similarity matrix, you can use the function read.matrix if you want to read the matrix from a file, or you can compute it using the function compute.matrix. In this case we load the data Sim1 and Sim2, which are included in the package, and compute a distance similarity matrix, with the function compute.matrix.

data(Sim1)
data(Sim2)
Dis1 = compute.matrix(net1=hpy)
Dis2 = compute.matrix(net1=sce)

Sim1 = (Sim1+Dis1)/2
Sim2 = (Sim2+Dis2)/2

When we have the networks and the similarity matrices, we can compute the clusters with the functions cluster.network and extract.clusters, and then visualize the clusters with display.clusters.

clust1 = cluster.network(sigma=Sim1,lambda=0.2,k=5)

clusters1 = extract.clusters(Net=hpy,ClustMat=clust1)


for(i in 1:length(clusters1)){
  print(names(clusters1)[i])
  plot(clusters1[[i]])
}

par(oma = c(3,3,4,3))
display.clusters(clust = clust1, Net = hpy, main = "")
cols = c("yellow", "black", "red", "green")
 legend(x = 0, y = 1.25, legend = 0:3, fill = cols, horiz = TRUE, bty = "n", xpd = TRUE, cex = 3)

clust2 = cluster.network(sigma = Sim2, lambda = 0.2, k = 5)

clusters2 = extract.clusters(Net = sce,ClustMat = clust2)


for (i in 1:length(clusters2)) {
  print(names(clusters2)[i])
  plot(clusters2[[i]])
}

par(oma = c(3,3,4,3))
display.clusters(clust = clust2,Net = sce,main = "")
cols = c("yellow","black","red","green")
 legend(x = 0, y = 1.25, legend = 0:3, fill = cols, horiz = TRUE, bty = "n", xpd = TRUE, cex = 3)

Once we have the clusters, we can compute the local alignments using align.local.all. We can use a dissimilarity matrices to compute the local alignments.

data(Sim)
localAligns = align.local.all(clust1 = clusters1,clust2 = clusters2, mat = Sim, threshold = 0)
localAligns2 = align.local.all(clust1 = clusters1, clust2 = clusters2, mat = Sim, threshold = 0, cores = 1, dismat = 1 - Sim)

Finally, using the local alignments, we can compute the global alignment, using align.global.

# scores = size.score.all(localAligns=localAligns)
# scores2 = sim.score.all(localAligns=localAligns,sim=Sim)
# scores[,2] = as.numeric(scores[,2])/5+as.numeric(scores2[,2])

alinGlobal = align.global(localAligns=localAligns,Sim=Sim)
alinGlobal2 = align.global(localAligns=localAligns2,Sim=Sim)

To compute the edge correctness score of the global alignment, we use the function EC.score.

EC.score(alin = alinGlobal, net1 = hpy, net2 = sce)
EC.score(alin = alinGlobal2, net1 = hpy, net2 = sce)

To compute the functional coherence score of the global alignment, we use the function FC.score.

To visualize the global alignment, we can plot all the alignment with align.plot, or visualize only one local alignment using align.local.plot. The first one is useful when we have small networks, and the second one for large networks.

align.plot(net1 = hpy,net2 = sce,global = alinGlobal2,k1 = 1,k2 = 1,edge.curved = 0.5,vertex.size = 5)

par(oma = c(1,2,1,2))
p1 = "85962.HP0303"
p2 = "RL2A_YEAST"
align.local.plot(localAligns = localAligns, global = alinGlobal2, p1 = p1, p2 = p2, net1 = hpy, net2 = sce)

par(oma = c(1,2,1,2))

p1 = "85962.HP1316"
p2 = "HAS1_YEAST"
align.local.plot(localAligns = localAligns, global = alinGlobal, p1 = p1, p2 = p2, net1 = hpy, net2 = sce)

RicUIB/AligNet documentation built on May 28, 2019, 2:27 p.m.