AKROUCHE Wassila
FERRAH Hichem
KACEL Lisa

rm(list = ls(all=TRUE))
library(miic)
library(bnlearn)
library(pcalg)

## 
## Attachement du package : 'pcalg'

## Les objets suivants sont masqués depuis 'package:bnlearn':
## 
##     dsep, pdag2dag, shd, skeleton

library(igraph)

## 
## Attachement du package : 'igraph'

## Les objets suivants sont masqués depuis 'package:bnlearn':
## 
##     as.igraph, compare, degree, subgraph

## Les objets suivants sont masqués depuis 'package:stats':
## 
##     decompose, spectrum

## L'objet suivant est masqué depuis 'package:base':
## 
##     union

library(qgraph)

1. Network reconstruction with the hill-climbing approach

1.1 Loading and exploring the dataset

data("cosmicCancer")
head(cosmicCancer)

dim(cosmicCancer)

## [1] 807 176

1.2 Calling the hill-climbing approach from the bnlearn package

En appelant l’approche hill climbing sur les données initiales nous faisons face à deux problématiques : - La présence de lignes vides NA - La présence de colonnes constantes Afin de pouvoir appliquer l’approche HC nous supprimons ces dernières ce qui résout le problème.

cosmicCancer_hc = bnlearn::hc(cosmicCancer)

## Error in check.data(x): the data set contains NaN/NA values.

1.3 Removing the samples for which at least one variable has a NA value

na_rows = cosmicCancer[!complete.cases(cosmicCancer),]
na_idx = as.integer(rownames(na_rows))
cosmicCancer_noNA = cosmicCancer[-na_idx,]
dim(cosmicCancer_noNA)

## [1] 799 176

cosmicCancer_noNA[, "Ploidy"] = as.factor(cosmicCancer_noNA[, "Ploidy"])

cosmicCancer_hc = bnlearn::hc(cosmicCancer_noNA)

## Error in check.data(x): variable esm1 must have at least two levels.

const_col = c()
for(iCol in c(1:ncol(cosmicCancer_noNA))){
  
  if(length(unique(as.character(cosmicCancer_noNA[, iCol]))) < 2){
    const_col = c(const_col, iCol) 
  }
}

cosmicCancer_noNACst = cosmicCancer_noNA[, -const_col]
cosmicCancer_hc = bnlearn::hc(cosmicCancer_noNACst)

cosmicCancer_hc_adj = amat(cosmicCancer_hc)

1.4 Making an igraph from the adjacency matrix

cosmicCancer_hc_net = igraph::graph_from_adjacency_matrix(cosmicCancer_hc_adj)
plot(cosmicCancer_hc_net, edge.arrow.size = .3,
     edge.color = "orange", vertex.color = "orange",
     vertex.frame.color ="#FFFFFF",
     vertex.label.color = "black", vertex.label.cex = .6,
     vertex.label = colnames(cosmicCancer),
     layout = layout.fruchterman.reingold)

Removing 0 degrees

Isolated = which(igraph::degree(cosmicCancer_hc_net)==0)
cosmicCancer_hc_net_wo0d = delete.vertices(cosmicCancer_hc_net, Isolated)

Plotting new graph

e <- get.edgelist(cosmicCancer_hc_net_wo0d,names=FALSE)
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(cosmicCancer_hc_net_wo0d),
                                       area=8*(vcount(cosmicCancer_hc_net_wo0d)^2),repulse.rad=(vcount(cosmicCancer_hc_net_wo0d)^3.1))

Coloring the genes

# Coloriage des genes comme spécifié 

cosmicCancer_names <- V(cosmicCancer_hc_net_wo0d)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"green","red")
cosmicCancer_colors[143]<- "yellow"

Graph displaying

#Affichage du Graphe

plot(cosmicCancer_hc_net_wo0d, edge.arrow.size = .3,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .5,
     vertex.label = cosmicCancer_names,vertex.size = 10,
     layout = l2)

## 1.5 Mutated Genes that are significantly related to gene expression

cosmicCancer_maj <- cosmicCancer_hc_adj[which(grepl("^[[:upper:]]+$", substr(rownames(cosmicCancer_hc_adj),1,1))),]
cosmicCancer_maj <- cosmicCancer_maj[,-which(grepl("^[[:upper:]]+$", substr(colnames(cosmicCancer_maj),1,1)))]

mutated_genes <- sort(colSums(cosmicCancer_maj),decreasing = TRUE)
mutated_genes[1:10]

##   tp53    rb1 scube2   flt1 rassf7   gnaz   mtdh   lin9 nusap1 rundc1 
##      3      2      1      1      1      1      1      1      1      1

Number of Relations that includes “Ploidy” gene

sum_relations_rows <- rowSums(cosmicCancer_hc_adj)
sum_relations_cols <- colSums(cosmicCancer_hc_adj)

ploidy_relations <- sum_relations_rows['Ploidy'] + sum_relations_cols['Ploidy']
print(ploidy_relations)

## Ploidy 
##      3

Top 10 nodes in terms of betweenness centrality measure

vertex_betws <- sort(betweenness(
  cosmicCancer_hc_net_wo0d,
  directed = TRUE,
  weights = NULL,
  nobigint = TRUE
),decreasing = TRUE)

vertex_betws[1:10]

##     tp53     MCM6      dck      rb1    chek1      bax    CDCA7    NDRG1 
## 424.1667 332.0000 300.0000 299.0000 294.0000 256.0000 252.7333 240.0000 
##   DIAPH3   gpr126 
## 232.0000 221.0000

Top 10 edges in terms of betweenness centrality measure

edges = as_edgelist(cosmicCancer_hc_net_wo0d)
edges_names = paste(edges[,1], edges[,2], sep="->")

edge_betws <- edge_betweenness(cosmicCancer_hc_net_wo0d, 
                               e = E(cosmicCancer_hc_net_wo0d), 
                               directed = TRUE, 
                               weights = NULL)

names(edge_betws) <- edges_names

edge_betws <- sort(edge_betws,decreasing = T)

edge_betws[1:10]

##     chek1->dck     rb1->chek1    CDCA7->MCM6       dck->bax    bax->gpr126 
##          315.0          308.0          300.5          272.0          238.0 
## GPR180->DIAPH3    AURKA->tp53      MCM6->rb1    RB1->GPR180  DIAPH3->NDRG1 
##          235.0          230.5          212.0          209.0          191.5

2. Network reconstruction with the PC approach

data(cosmicCancer)

2.2 Calling the PC approach

- PC - ALPHA = 0.01

cosmicCancer_noNACst <- Filter(function(x)(length(unique(x))>1), cosmicCancer_noNACst)

cosmicCancer_pc_data = data.matrix(cosmicCancer_noNACst)
cosmicCancer_pc_data = cosmicCancer_pc_data-1
n_lev = rep(-1, ncol(cosmicCancer_pc_data))
for(iCol in 1:length(n_lev)){
  n_lev[iCol] = length(unique(cosmicCancer_pc_data[,iCol]))
}
suffStat <- list(dm = cosmicCancer_pc_data, nlev = n_lev, adaptDF = FALSE)
pc.cancer_data <- pc(suffStat, indepTest = disCItest,
                     p = ncol(cosmicCancer_noNACst), alpha = 0.01,
                     verbose = TRUE)
pc.cancer_data_bn = as.bn(pc.cancer_data, check.cycles = FALSE)

pc.cancer_net = igraph::graph_from_adjacency_matrix(amat(pc.cancer_data_bn))

# suppression des sommets de degré 0

Isolated = which(igraph::degree(pc.cancer_net)==0)
V(pc.cancer_net)$name <- colnames(cosmicCancer_pc_data)
pc.cancer_net_wo0d = delete.vertices(pc.cancer_net, Isolated)

# Amelioration d'affichages ( sans overlapping )

e <- get.edgelist(pc.cancer_net_wo0d,names=FALSE)
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(pc.cancer_net_wo0d),
                                        area=8*(vcount(pc.cancer_net_wo0d)^2),repulse.rad=(vcount(pc.cancer_net_wo0d)^3.1))

# Coloriage des genes comme spécifié 

cosmicCancer_names <- V(pc.cancer_net_wo0d)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"green","red")
#length(cosmicCancer_names)
#cosmicCancer_names[63]
cosmicCancer_colors[63]<- "yellow"
#Affichage du Graphe

plot(pc.cancer_net_wo0d, edge.arrow.size = .3,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .5,
     vertex.label = cosmicCancer_names,vertex.size = 10,
     layout = l2)

# Mutated Genes that are significantly related to gene expression 
pc.cancer_mat <- amat(pc.cancer_data_bn)
colnames(pc.cancer_mat) <- colnames(cosmicCancer_pc_data)
rownames(pc.cancer_mat) <- colnames(cosmicCancer_pc_data)

cosmicCancer_maj <- pc.cancer_mat[which(grepl("^[[:upper:]]+$", substr(rownames(pc.cancer_mat),1,1))),]
cosmicCancer_maj <- cosmicCancer_maj[,-which(grepl("^[[:upper:]]+$", substr(colnames(pc.cancer_mat),1,1)))]

mutated_genes <- sort(colSums(cosmicCancer_maj),decreasing = TRUE)
mutated_genes[1:10]

##   bbc3  egln1  tgfb3 igfbp5  fgf18 scube2  wisp1   flt1 hrasls stk32b 
##      0      0      0      0      0      0      0      0      0      0

# Number of Relations that includes "Ploidy" gene

sum_relations_rows <- rowSums(pc.cancer_mat)
sum_relations_cols <- colSums(pc.cancer_mat)

ploidy_relations <- sum_relations_rows['Ploidy'] + sum_relations_cols['Ploidy']
print(ploidy_relations)

## Ploidy 
##      2

# Top 10 nodes in terms of betweenness centrality measure
vertex_betws <- sort(betweenness(
  pc.cancer_net_wo0d,
  directed = TRUE,
  weights = NULL,
  nobigint = TRUE
),decreasing = TRUE)

vertex_betws[1:10]

##   PRC1 DIAPH3   GMPS  FOXM1   ECT2  BRCA2 GPR180    RB1   RFC4   LIN9 
##     93     88     83     76     52     46     28     17     15     11

# Top 10 edges in terms of betweenness centrality measure

edges = as_edgelist(pc.cancer_net_wo0d)
edges_names = paste(edges[,1], edges[,2], sep="->")

edge_betws <- edge_betweenness(pc.cancer_net_wo0d, 
                               e = E(pc.cancer_net_wo0d), 
                               directed = TRUE, 
                               weights = NULL)

names(edge_betws) <- edges_names

edge_betws <- sort(edge_betws,decreasing = T)

edge_betws[1:10]

##   PRC1->DIAPH3    FOXM1->PRC1    GMPS->FOXM1     ECT2->GMPS    BRCA2->ECT2 
##            103             86             77             62             56 
## DIAPH3->GPR180  DIAPH3->BRCA2    GPR180->RB1     RFC4->GMPS    LIN9->BRCA2 
##             44             39             30             26             22

#### ALPHA = 0.005


pc.cancer_data <- pc(suffStat, indepTest = disCItest,
                     p = ncol(cosmicCancer_noNACst), alpha = 0.005,
                     verbose = TRUE)
pc.cancer_data_bn = as.bn(pc.cancer_data, check.cycles = FALSE)

pc.cancer_net = igraph::graph_from_adjacency_matrix(amat(pc.cancer_data_bn))

# suppression des sommets de degré 0

Isolated = which(igraph::degree(pc.cancer_net)==0)
V(pc.cancer_net)$name <- colnames(cosmicCancer_pc_data)
pc.cancer_net_wo0d = delete.vertices(pc.cancer_net, Isolated)

# Amelioration d'affichages ( sans overlapping )

e <- get.edgelist(pc.cancer_net_wo0d,names=FALSE)
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(pc.cancer_net_wo0d),
                                        area=8*(vcount(pc.cancer_net_wo0d)^2),repulse.rad=(vcount(pc.cancer_net_wo0d)^3.1))

# Coloriage des genes comme spécifié 

cosmicCancer_names <- V(pc.cancer_net_wo0d)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"green","red")

#cosmicCancer_names[length(cosmicCancer_names)]
cosmicCancer_colors[length(cosmicCancer_names)]<- "yellow"

#Affichage du Graphe

plot(pc.cancer_net_wo0d, edge.arrow.size = .3,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .5,
     vertex.label = cosmicCancer_names,vertex.size = 10,
     layout = l2)

# Mutated Genes that are significantly related to gene expression 
pc.cancer_mat <- amat(pc.cancer_data_bn)
colnames(pc.cancer_mat) <- colnames(cosmicCancer_pc_data)
rownames(pc.cancer_mat) <- colnames(cosmicCancer_pc_data)

cosmicCancer_maj <- pc.cancer_mat[which(grepl("^[[:upper:]]+$", substr(rownames(pc.cancer_mat),1,1))),]
cosmicCancer_maj <- cosmicCancer_maj[,-which(grepl("^[[:upper:]]+$", substr(colnames(pc.cancer_mat),1,1)))]

mutated_genes <- sort(colSums(cosmicCancer_maj),decreasing = TRUE)
mutated_genes[1:10]

##   bbc3  egln1  tgfb3 igfbp5  fgf18 scube2  wisp1   flt1 hrasls stk32b 
##      0      0      0      0      0      0      0      0      0      0

# Number of Relations that includes "Ploidy" gene

sum_relations_rows <- rowSums(pc.cancer_mat)
sum_relations_cols <- colSums(pc.cancer_mat)

ploidy_relations <- sum_relations_rows['Ploidy'] + sum_relations_cols['Ploidy']
print(ploidy_relations)

## Ploidy 
##      2

# Top 10 nodes in terms of betweenness centrality measure
vertex_betws <- sort(betweenness(
  pc.cancer_net_wo0d,
  directed = TRUE,
  weights = NULL,
  nobigint = TRUE
),decreasing = TRUE)

vertex_betws[1:10]

## DIAPH3   ECT2   MCM6  CDCA7    RB1 GPR180  BRCA2  AURKA  CCNE2 NUSAP1 
##      8      6      6      6      6      4      4      4      3      3

# Top 10 edges in terms of betweenness centrality measure

edges = as_edgelist(pc.cancer_net_wo0d)
edges_names = paste(edges[,1], edges[,2], sep="->")

edge_betws <- edge_betweenness(pc.cancer_net_wo0d, 
                               e = E(pc.cancer_net_wo0d), 
                               directed = TRUE, 
                               weights = NULL)

names(edge_betws) <- edges_names

edge_betws <- sort(edge_betws,decreasing = T)

edge_betws[1:10]

##    CDCA7->MCM6     ECT2->GMPS    RB1->CDKN2A      MCM6->RB1    GPR180->RB1 
##              8              7              7              6              6 
## DIAPH3->GPR180  DIAPH3->BRCA2    BRCA2->ECT2   CDCA7->FOXM1 GPR180->DIAPH3 
##              6              6              6              5              5

### ALPHA = 0.001



pc.cancer_data <- pc(suffStat, indepTest = disCItest,
                     p = ncol(cosmicCancer_noNACst), alpha = 0.001,
                     verbose = TRUE)
pc.cancer_data_bn = as.bn(pc.cancer_data, check.cycles = FALSE)

pc.cancer_net = igraph::graph_from_adjacency_matrix(amat(pc.cancer_data_bn))

# suppression des sommets de degré 0

Isolated = which(igraph::degree(pc.cancer_net)==0)
V(pc.cancer_net)$name <- colnames(cosmicCancer_pc_data)
pc.cancer_net_wo0d = delete.vertices(pc.cancer_net, Isolated)

# Amelioration d'affichages ( sans overlapping )

e <- get.edgelist(pc.cancer_net_wo0d,names=FALSE)
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(pc.cancer_net_wo0d),
                                        area=8*(vcount(pc.cancer_net_wo0d)^2),repulse.rad=(vcount(pc.cancer_net_wo0d)^3.1))

# Coloriage des genes comme spécifié 

cosmicCancer_names <- V(pc.cancer_net_wo0d)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"green","red")
#cosmicCancer_names
#No Ploidy

#Affichage du Graphe

plot(pc.cancer_net_wo0d, edge.arrow.size = .3,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .5,
     vertex.label = cosmicCancer_names,vertex.size = 10,
     layout = l2)

# Mutated Genes that are significantly related to gene expression 
pc.cancer_mat <- amat(pc.cancer_data_bn)
colnames(pc.cancer_mat) <- colnames(cosmicCancer_pc_data)
rownames(pc.cancer_mat) <- colnames(cosmicCancer_pc_data)

cosmicCancer_maj <- pc.cancer_mat[which(grepl("^[[:upper:]]+$", substr(rownames(pc.cancer_mat),1,1))),]
cosmicCancer_maj <- cosmicCancer_maj[,-which(grepl("^[[:upper:]]+$", substr(colnames(pc.cancer_mat),1,1)))]

mutated_genes <- sort(colSums(cosmicCancer_maj),decreasing = TRUE)
mutated_genes[1:10]

##   bbc3  egln1  tgfb3 igfbp5  fgf18 scube2  wisp1   flt1 hrasls stk32b 
##      0      0      0      0      0      0      0      0      0      0

# Number of Relations that includes "Ploidy" gene

sum_relations_rows <- rowSums(pc.cancer_mat)
sum_relations_cols <- colSums(pc.cancer_mat)

ploidy_relations <- sum_relations_rows['Ploidy'] + sum_relations_cols['Ploidy']
print(ploidy_relations)

## Ploidy 
##      0

# Top 10 nodes in terms of betweenness centrality measure
vertex_betws <- sort(betweenness(
  pc.cancer_net_wo0d,
  directed = TRUE,
  weights = NULL,
  nobigint = TRUE
),decreasing = TRUE)

vertex_betws[1:10]

##   ECT2   GMPS   RFC4  CDCA7  AURKB  EGLN1  CENPA   MCM6  FOXM1 GPR180 
##      8      8      5      4      3      2      2      2      2      1

# Top 10 edges in terms of betweenness centrality measure

edges = as_edgelist(pc.cancer_net_wo0d)
edges_names = paste(edges[,1], edges[,2], sep="->")

edge_betws <- edge_betweenness(pc.cancer_net_wo0d, 
                               e = E(pc.cancer_net_wo0d), 
                               directed = TRUE, 
                               weights = NULL)

names(edge_betws) <- edges_names

edge_betws <- sort(edge_betws,decreasing = T)

edge_betws[1:10]

##   GMPS->RFC4   ECT2->GMPS RFC4->HRASLS  CCNE2->ECT2 NUSAP1->ECT2  CDCA7->MCM6 
##           10            9            6            5            5            4 
## CDCA7->FOXM1 AURKB->CENPA   tp53->GMPS  ECT2->BRCA2 
##            4            4            3            3

3.Network reconstruction with the MIIC approach

data(cosmicCancer)

3.1.build the cosmicCancer network

with MIIC.

# execute MIIC (reconstruct graph)
miic.res <- miic(
input_data = cosmicCancer, state_order = cosmicCancer_stateOrder, latent = "yes",
n_shuffles = 100, conf_threshold = 0.001
)

# plot graph
if(require(igraph)) {
  
plot(miic.res)
}

3.2.2

confidenceShuffle : représente le nombre de brassages du dataset originale dans le but d’évaluer le rapport de confiance spécifique aux bords de tous les bords inférés.

confidenceThreshold : Le seuil utilisé pour filtrer les arêtes les moins probables après l’étape du squelette.

# getting several network with different values of threshold and shuffles
miic.res1 <- miic(
input_data = cosmicCancer, state_order = cosmicCancer_stateOrder, latent = "yes",
n_shuffles = 200, conf_threshold = 0.025
)
miic.res2 <- miic(
input_data = cosmicCancer, state_order = cosmicCancer_stateOrder, latent = "yes",
n_shuffles = 50, conf_threshold = 0.1
)
miic.res3 <- miic(
input_data = cosmicCancer, state_order = cosmicCancer_stateOrder, latent = "yes",
n_shuffles = 50, conf_threshold = 0.001
)

3.2.3 Convert the MIIC network to an igraph object and plot the result.

miic_res_net = igraph::graph_from_adjacency_matrix(miic.res$adj_matrix,
                                                       mode="directed")

miic_res_net1 = igraph::graph_from_adjacency_matrix(miic.res1$adj_matrix,
                                                       mode="directed")
miic_res_net2 = igraph::graph_from_adjacency_matrix(miic.res2$adj_matrix,
                                                       mode="directed")
miic_res_net3 = igraph::graph_from_adjacency_matrix(miic.res3$adj_matrix,
                                                       mode="directed")

Plotting the igraph without removing 0 degrees

plot(miic_res_net, edge.arrow.size = .3, 
     edge.color = "orange", vertex.color = "orange",
     vertex.frame.color ="#ffffff", edge.curved = 0,
     vertex.label.color = "black", vertex.label.cex = .6,
     vertex.label = colnames(cosmicCancer),
     layout = layout.fruchterman.reingold)

## Removing 0 degrees variables and coloring differently

Isolated = which(igraph::degree(miic_res_net)==0)
Isolated1 = which(igraph::degree(miic_res_net1)==0)
Isolated2 = which(igraph::degree(miic_res_net2)==0)
Isolated3 = which(igraph::degree(miic_res_net3)==0)
miic_res_net_wo0d = delete.vertices(miic_res_net, Isolated)
miic_res_net_wo0d1 = delete.vertices(miic_res_net, Isolated1)
miic_res_net_wo0d2 = delete.vertices(miic_res_net, Isolated2)
miic_res_net_wo0d3 = delete.vertices(miic_res_net, Isolated3)

Coloring and displaying the different graphs

# Coloriage des genes comme spécifié 
cosmicCancer_names <- V(miic_res_net_wo0d)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"purple","green")
#length(cosmicCancer_names)
#cosmicCancer_colors[69]<- "yellow"
e <- get.edgelist(miic_res_net_wo0d,names=FALSE)
l <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d))
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d),
                                       area=8*(vcount(miic_res_net_wo0d1)^2),repulse.rad=(vcount(miic_res_net_wo0d)^3.1))


plot(miic_res_net_wo0d, edge.arrow.size = .1,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .4,
     vertex.label = colnames(cosmicCancer),vertex.size = 10,
     main="n_shuffles = 100 , threshold = 0.001 (from the doc) ",
     layout = l2)

cosmicCancer_names <- V(miic_res_net_wo0d1)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"purple","green")
#length(cosmicCancer_names)
cosmicCancer_colors[117]<- "yellow"
e <- get.edgelist(miic_res_net_wo0d1,names=FALSE)
l <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d1))
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d1),
                                       area=8*(vcount(miic_res_net_wo0d1)^2),repulse.rad=(vcount(miic_res_net_wo0d1)^3.1))


plot(miic_res_net_wo0d1, edge.arrow.size = .1,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .4,
     vertex.label = colnames(cosmicCancer),vertex.size = 10,
     main="n_shuffles = 200 , threshold = 0.025 ",
     layout = l2)

cosmicCancer_names <- V(miic_res_net_wo0d2)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"purple","green")
#length(cosmicCancer_names)
cosmicCancer_colors[141]<- "yellow"

e <- get.edgelist(miic_res_net_wo0d2,names=FALSE)
l <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d2))
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d2),
                                       area=8*(vcount(miic_res_net_wo0d2)^2),repulse.rad=(vcount(miic_res_net_wo0d2)^3.1))


plot(miic_res_net_wo0d2, edge.arrow.size = .1,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .4,
     vertex.label = colnames(cosmicCancer),vertex.size = 10,
     main="n_shuffles = 50 , threshold = 0.1 ",
     layout = l2)

cosmicCancer_names <- V(miic_res_net_wo0d3)$name
cosmicCancer_colors <- ifelse(grepl("^[[:upper:]]+$", substr(cosmicCancer_names,1,1)),"purple","green")
#length(cosmicCancer_names)
cosmicCancer_colors[143]<- "yellow"
e <- get.edgelist(miic_res_net_wo0d3,names=FALSE)
l <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d3))
l2 <- qgraph.layout.fruchtermanreingold(e,vcount=vcount(miic_res_net_wo0d3),
                                       area=8*(vcount(miic_res_net_wo0d3)^2),repulse.rad=(vcount(miic_res_net_wo0d3)^3.1))


plot(miic_res_net_wo0d3, edge.arrow.size = .1,
     edge.color = "blue", vertex.color = cosmicCancer_colors,
     vertex.frame.color ="blue",
     vertex.label.color = "black", vertex.label.cex = .4,
     vertex.label = colnames(cosmicCancer),vertex.size = 10,
     main="n_shuffles = 50 , threshold = 0.001 ",
     layout = l2)

1.4 Mutated Genes that are significantly related to gene expression

micres_maj <- miic.res$adj_matrix[which(grepl("^[[:upper:]]+$", substr(rownames(miic.res$adj_matrix),1,1))),]
micres_maj <- micres_maj[,-which(grepl("^[[:upper:]]+$", substr(colnames(micres_maj),1,1)))]

mutated_genes <- sort(colSums(micres_maj),decreasing = TRUE)
mutated_genes[1:10]

##    rb1   bbc3  egln1  tgfb3   esm1 igfbp5  fgf18 scube2  wisp1   flt1 
##      2      0      0      0      0      0      0      0      0      0

Number of Relations that includes “Ploidy” gene

sum_relations_rows <- rowSums(miic.res$adj_matrix)
sum_relations_cols <- colSums(miic.res$adj_matrix)

ploidy_relations <- sum_relations_rows['Ploidy'] + sum_relations_cols['Ploidy']
print(ploidy_relations)

## Ploidy 
##      2

Top 10 nodes in terms of betweenness centrality measure

vertex_betws <- sort(betweenness(
  miic_res_net_wo0d,
  directed = TRUE,
  weights = NULL,
  nobigint = TRUE
),decreasing = TRUE)

vertex_betws[1:10]

##      tp53     CENPA      GMPS     FOXM1     AURKA      ECT2     CCNE2     NDRG1 
## 486.33333 379.00000 310.66667 246.50000 227.66667 175.00000 158.00000  97.00000 
##    DIAPH3     CHEK1 
##  70.66667  53.33333

Top 10 edges in terms of betweenness centrality measure

edges = as_edgelist(miic_res_net_wo0d)
edges_names = paste(edges[,1], edges[,2], sep="->")

edge_betws <- edge_betweenness(miic_res_net_wo0d, 
                               e = E(miic_res_net_wo0d), 
                               directed = TRUE, 
                               weights = NULL)

names(edge_betws) <- edges_names

edge_betws <- sort(edge_betws,decreasing = T)

edge_betws[1:10]

##  CENPA->tp53  CENPA->tp53   tp53->GMPS   tp53->GMPS AURKA->CENPA AURKA->CENPA 
##    148.66667    148.66667    142.00000    142.00000    139.83333    139.83333 
##  ECT2->CCNE2  ECT2->CCNE2  GMPS->FOXM1  GMPS->FOXM1 
##     87.50000     87.50000     79.08333     79.08333

TP évalué : Analysis of genomic and ploidy alterations in breast tumors

1. Network reconstruction with the hill-climbing approach

1.1 Loading and exploring the dataset

1.2 Calling the hill-climbing approach from the bnlearn package

1.3 Removing the samples for which at least one variable has a NA value

1.4 Making an igraph from the adjacency matrix

Removing 0 degrees

Plotting new graph

Coloring the genes

Graph displaying

Number of Relations that includes “Ploidy” gene

Top 10 nodes in terms of betweenness centrality measure

Top 10 edges in terms of betweenness centrality measure

2. Network reconstruction with the PC approach

2.2 Calling the PC approach

- PC - ALPHA = 0.01

3.Network reconstruction with the MIIC approach

3.1.build the cosmicCancer network

3.2.2

3.2.3 Convert the MIIC network to an igraph object and plot the result.

Plotting the igraph without removing 0 degrees

Coloring and displaying the different graphs

1.4 Mutated Genes that are significantly related to gene expression

Number of Relations that includes “Ploidy” gene

Top 10 nodes in terms of betweenness centrality measure

Top 10 edges in terms of betweenness centrality measure