-
Notifications
You must be signed in to change notification settings - Fork 0
/
02_Survival_Analysis.R
171 lines (139 loc) · 5.83 KB
/
02_Survival_Analysis.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
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
#######################################
####### Survival Analysis ##
#######################################
# TODO: Clean data
# TODO: Figure out what time to event data to use. Both the patient and
# the follow up clinical info have slightly different days to death data
# Also I think there's some progression free time data hiding somewhere
# TODO: Determine clinical variables for inclusion in the Cox model
# TODO: Get code ready
#######################################
######### Libraries ############
#######################################
library("TCGAbiolinks")
library(survival)
#######################################
######### Data Preparation ######
#######################################
#Load the clinical data
clinical.query <- GDCquery(project = "TCGA-OV",
data.category = "Clinical")
# TODO: Clean, figure out what's useful
clinical <- GDCprepare_clinic(clinical.query, clinical.info = "patient")
clinical.drug <- GDCprepare_clinic(clinical.query, clinical.info = "drug")
clinical.fup <- GDCprepare_clinic(clinical.query, clinical.info = "follow_up")
clinical.admin <- GDCprepare_clinic(clinical.query, clinical.info = "admin")
#Might want to only choose popular drugs
sort(table(clinical.drug$drug_name))
#INFORMATION FROMhttps://pubchem.ncbi.nlm.nih.gov/compound/paclitaxel#section=Top
# Some of the popular drugs (only 50+ uses)
# Carboplatin IS platinum based
# Taxol IS NOT
# paclitaxel IS NOT
# cisplatin IS platinum based
# Doxil IS NOT
# Topotecan IS NOT
# Taxotere IS NOT
# Gemcitabine IS NOT
# Gemzar IS NOT
plat <- c("Carboplatin","Cisplatin")
#However, most times chemotherapy drugs are given in combos so thats the next thing to do, sort by platinum vs not use
#And Carboplatin vs Cisplatin
#Plat vs non_plat
patient_names <- names(table(clinical.drug$bcr_patient_barcode))
plat_use <- rep(NA,length(patient_names))
plat_type <- rep(NA,length(patient_names))
for (i in c(1: length(patient_names)))
{
data <- clinical.drug[clinical.drug$bcr_patient_barcode == patient_names[i],]
if (any(plat %in% data$drug_name) == "TRUE"){
plat_use[i] <- 1
}
else{
plat_use[i] <- 0
plat_type[i] <- "Neither"
}
if (plat[1] %in% data$drug_name & plat[2] %in% data$drug_name){
plat_type[i] <- "Both"
}
else if (plat[1] %in% data$drug_name & !(plat[2] %in% data$drug_name)){
plat_type[i] <- plat[1]
}
else if (!(plat[1] %in% data$drug_name) & plat[2] %in% data$drug_name){
plat_type[i] <- plat[2]
}
else{
plat_type[i] <- "Neither"
}
}
temp1 <- data.frame("bcr_patient_barcode" = patient_names,plat_use,plat_type)
clinical <- merge(clinical, temp1, by="bcr_patient_barcode")
plat_use2 <- plat_use
plat_type2 <- plat_type
plat_use <- c()
plat_type <- c()
for (i in 1:length(patient_names)){
data <- clinical.drug[clinical.drug$bcr_patient_barcode == patient_names[i],]
reps <- dim(data)[1]
if (plat_use2[i] == 1){
plat_use <- c(plat_use, rep(1,reps))
}
else{
plat_use <- c(plat_use, rep(0,reps))
}
if (plat_type2[i] == "Both"){
plat_type <- c(plat_type, rep("Both", reps))
}
else if (plat_type2[i] == plat[1]){
plat_type <- c(plat_type, rep(plat[1], reps))
}
else if (plat_type2[i] == plat[2]){
plat_type <- c(plat_type, rep(plat[2], reps))
}
else{
plat_type <- c(plat_type, rep("Neither", reps))
}
}
clinical.drug <- clinical.drug
temp <- data.frame(plat_use, plat_type)
clinical.drug <- cbind(clinical.drug, temp)
#Roughly, 0.9396226 of the patients were treated with platinums. Might be enought to do a comparision study?
#Rougly 0.06 were treated with only cisplatin
#So it seems that we have decent numbers for comparsion between platinum based and wihotut platinum studiess(al least a perliminary)
#######################################
######### Kaplan Meier ######
#######################################
#Important variables
str(clinical)
str(clinical.drug)
str(clinical.fup)
surv.vars <- c("bcr_patient_barcode", "days_to_death", "days_to_last_followup",
"age_at_initial_pathologic_diagnosis", "race_list", "person_neoplasm_cancer_status",
"ethnicity","neoplasm_histologic_grade", "radiation_therapy", "primary_therapy_outcome_success", "ethnicity",
"plat_use", "plat_type")
surv.drug.vars <- c("bcr_drug_barcode", "therapy_types", "drug_name", "regimen_indication", "plat_use", "plat_type")
surv.data <- clinical[, surv.vars]
surv.drug.data <- clinical.drug[, surv.drug.vars]
# Define the time to event. If we have the days to death, use that.
# Otherwise use their last follow up
surv.data$EventTime <- ifelse(!is.na(surv.data$days_to_death),
surv.data$days_to_death,
surv.data$days_to_last_followup)
# Create an indicator variable for whether the event is death or censoring
surv.data$censored <- ifelse(!is.na(surv.data$days_to_death),
FALSE,
TRUE)
# Fit the KM curves
km.surv.fit <- survfit(Surv(surv.data$EventTime, surv.data$censored) ~ surv.data$plat_type , conf.type = "plain")
# Example KM Plot
plot(km.surv.fit, main=expression(paste("Kaplan-Meier-estimate ", hat(S)[g](t), " for Different drug use")),
xlab="t", ylab="Survival", lwd=2, col=1:4)
legend(x="topright", col=1:4, lwd=2, legend=c("Both", "Carboplatin Only", "Cisplatin Only", "Neither"))
#######################################
######### Cox Model ######
#######################################
# Cox model based on age. Just for example
cox.age <- coxph(data = surv.data, Surv(EventTime, censored) ~ age_at_initial_pathologic_diagnosis + race_list + neoplasm_histologic_grade + plat_type + radiation_therapy + person_neoplasm_cancer_status)
cox.age <- coxph(data = surv.data, Surv(EventTime, censored) ~ age_at_initial_pathologic_diagnosis + radiation_therapy)
#Why am i getting nonconvergence?
summary(cox.age)