-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmain_withGO.R
127 lines (113 loc) · 5.24 KB
/
main_withGO.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
library(dplyr)
library(data.table)
epsilon <- 1e-10 # A very small number
getMatrix <- function(m,d1,d2){
m = matrix(m, nrow = length(d1), ncol = length(d2))
colnames(m) = d2
row.names(m) = d1
return(m)
}
# Read inputs -------------------------------------------------------------
drugs <- read.csv("./Data/external/drug links.csv")
drugs <- subset(drugs, Drug.Type == "SmallMoleculeDrug")
drugMetaData = read.csv("./Data/DB_full_meta_data.csv",header = TRUE,sep = ",", stringsAsFactors = FALSE)
drugs.ws <- drugMetaData[which(is.na(drugMetaData["Smiles"])),1] %>%
setdiff(drugs[,1], .)
drugs.ws.app <- intersect(drugs.ws, drugMetaData[grepl("approved",as.character(drugMetaData[,3])),1])
sc.m <- as.matrix(read.csv("./Data/chem_similarity.csv", row.names = 1))
sp.m <- as.matrix(read.csv("./Data/path_similarity_v2.csv", row.names = 1)) # improved: used BioCor package (for KEGG)
st.m <- as.matrix(read.csv("./Data/target_similarity.csv", row.names = 1))
sgoCC.m <- as.matrix(read.csv("./Data/GO_Sim/CC/GO_Sim_CC_combined.csv", row.names = 1))
sgoMF.m <- as.matrix(read.csv("./Data/GO_Sim/MF/GO_Sim_MF_combined.csv", row.names = 1))
sgoBP.m <- as.matrix(read.csv("./Data/GO_Sim/BP/GO_Sim_BP_combined.csv", row.names = 1))
# Remove self-similarities, sub-setting drug-pairs, -----------------------------
# denoising similarity matrices
# and finally Aggregate and p-value calculation -----------------------------
diag(sc.m) <- 0; diag(sp.m) <- 0; diag(st.m) <- 0; diag(sgoCC.m) <- 0; diag(sgoMF.m) <- 0; diag(sgoBP.m) <- 0;
sc.m <- sc.m[which(rownames(sc.m) %in% drugs.ws),]
sc.m <- sc.m[,which(colnames(sc.m) %in% drugs.ws.app)]
sp.m <- sp.m[which(rownames(sp.m) %in% drugs.ws),]
sp.m <- sp.m[,which(colnames(sp.m) %in% drugs.ws.app)]
st.m <- st.m[which(rownames(st.m) %in% drugs.ws),]
st.m <- st.m[,which(colnames(st.m) %in% drugs.ws.app)]
sgoCC.m <- sgoCC.m[which(rownames(sgoCC.m) %in% drugs.ws),]
sgoCC.m <- sgoCC.m[,which(colnames(sgoCC.m) %in% drugs.ws.app)]
sgoMF.m <- sgoMF.m[which(rownames(sgoMF.m) %in% drugs.ws),]
sgoMF.m <- sgoMF.m[,which(colnames(sgoMF.m) %in% drugs.ws.app)]
sgoBP.m <- sgoBP.m[which(rownames(sgoBP.m) %in% drugs.ws),]
sgoBP.m <- sgoBP.m[,which(colnames(sgoBP.m) %in% drugs.ws.app)]
# although it's less likely to have any NAs is n sc.m
# since we've discared drugs without any structures.
sc.m <- ifelse(is.na(sc.m), 0, sc.m)
# epsilon instead here just be fair for some drugs
# there may not have any pathway associated
# sp.m <- ifelse(is.na(sp.m), epsilon, sp.m)
# sp.m <- ifelse(is.na(sp.m), 0, sp.m)
st.m <- ifelse(is.na(st.m), 0, st.m)
sgoCC.m <- ifelse(is.na(sgoCC.m), epsilon, sgoCC.m)
sgoMF.m <- ifelse(is.na(sgoMF.m), epsilon, sgoMF.m)
sgoBP.m <- ifelse(is.na(sgoBP.m), epsilon, sgoBP.m)
m <- rowMeans(cbind(c(sc.m), c(st.m), c(sp.m), c(sgoCC.m), c(sgoMF.m), c(sgoBP.m)), na.rm = TRUE)
tmp <- m
tmp[tmp<0.05] = NA
p <- pnorm(scale(tmp),lower.tail = FALSE)
p.adj <- p.adjust(p, method = "fdr")
rm(tmp)
# Build the network
d1 = drugs.ws
d2 = drugs.ws.app
p = getMatrix(p, d1,d2)
p.adj = getMatrix(p.adj, d1, d2)
m = getMatrix(m,d1, d2)
# Output network ----------------------------------------------------------
# percentile = quantile(m, 0.99, na.rm = TRUE)
# library(foreach)
# library(doParallel)
# cl <- makeCluster(122)
# registerDoParallel(cl)
#
# out <- c()
# s <- Sys.time()
# out <- foreach(i = 1:length(d1), .combine='rbind') %:%
# foreach(j = 1:length(d2), .combine='rbind') %dopar% {
# di <- d1[i]
# dj <- d2[j]
# if(!is.na(m[di,dj]) & m[di,dj] > percentile){
# cbind(di,dj,sc.m[di,dj],st.m[di,dj],sp.m[di,dj], m[di,dj], p.adj[di,dj], p[di,dj])
# # r
# }
# }
# e <- Sys.time()
# print(e-s)
# out_perc <- c()
out_pval <- c()
s <- Sys.time()
counter <- 0
elapsed <- 0
for (i in d1) {
for (j in d2) {
# if(!is.na(m[i,j]) & m[i,j] > percentile){
# r <- c(i,j,sc.m[i,j],st.m[i,j],sp.m[i,j],sgoCC.m[i,j], sgoMF.m[i,j], sgoBP.m[i,j], m[i,j], p[i,j], p.adj[i,j])
# out_perc <- rbind(out_perc,r)
# }
if(!is.na(p[i,j]) & p[i,j] < 0.05){
r <- c(i,j,sc.m[i,j],st.m[i,j],sp.m[i,j], sgoCC.m[i,j], sgoMF.m[i,j], sgoBP.m[i,j], m[i,j], p[i,j], p.adj[i,j])
out_pval <- rbind(out_pval,r)
}
# if(!is.na(p[i,j])){
# r <- c(i,j,sc.m[i,j],st.m[i,j],sp.m[i,j], sgoCC.m[i,j], sgoMF.m[i,j], sgoBP.m[i,j], m[i,j], p[i,j], p.adj[i,j])
# out_pval <- rbind(out_pval,r)
# }
}
e <- Sys.time()
elapsed <- elapsed + (e-s)
print(paste0((counter/length(d1))*100, "% fininsed in: ", elapsed, " seconds: number of drug-pairs: ", nrow(out_pval)))
counter <- counter+1
}
# out1_perc <- as.data.frame(out_perc)
out1_pval <- as.data.frame(out_pval)
# colnames(out1_perc) <- c("ID1","ID2","Chem_similarity","Target_similarity","Pathway_similarity","GO_CC_Similarity", "GO_MF_Similarity","GO_BP_Similarity", "rowMeans","p_value","adjP_value")
colnames(out1_pval) <- c("ID1","ID2","Chem_similarity","Target_similarity","Pathway_similarity","GO_CC_Similarity", "GO_MF_Similarity","GO_BP_Similarity", "rowMeans","p_value","adjP_value")
# fwrite(out1_perc,"./Data/new_net_info_V4_perc.csv", na="NA", quote = F, row.names = F)
fwrite(out1_pval,"./Data/new_net_info_V7_pval.csv", na="NA", quote = F, row.names = F)
# write.csv(out, "new_net_info.csv",quote = F, row.names = F, col.names = F)