| Type: | Package |
| Title: | Personalized Disease Network |
| Version: | 0.1.0 |
| Author: | Javier Cabrera <xavier.cabrera@gmail.com>, Fei Wang <cherry8177@gmail.com> |
| Maintainer: | Zhenbang Wang <bruce1edward@gmail.com> |
| Description: | Building patient level networks for prediction of medical outcomes and draw the cluster of network. This package is based on paper Personalized disease networks for understanding and predicting cardiovascular diseases and other complex processes (See Cabrera et al. http://circ.ahajournals.org/content/134/Suppl_1/A14957). |
| License: | GPL-2 |
| Depends: | R (≥ 2.3.0) |
| Imports: | survival, glmnet, ggplot2, network |
| LazyData: | TRUE |
| RoxygenNote: | 6.0.1 |
| NeedsCompilation: | no |
| Packaged: | 2017-11-03 19:40:15 UTC; Bruce |
| VignetteBuilder: | knitr |
| Suggests: | knitr, rmarkdown |
| Repository: | CRAN |
| Date/Publication: | 2017-11-03 23:27:43 UTC |
Personized Disease Network
Description
Building patient level networks for prediction of medical outcomes and draw the cluster of network. This pakcage is based on paper Personalized disease networks (PDN) for understanding and predicting cardiovascular diseases and other complex processes.
Details
| Package: | PDN |
| Type: | Package |
| Version: | 1.0 |
| Date: | 2017-08-14 |
| License: | GPL-2 |
Author(s)
Javier Cabrera and Fei Wang
Maintainer: Zhenbang Wang <bruce1edward@gmail.com>
References
Cabrera, Javier and Kostis, John B and Wang, Fei and Kostis, William J. (2016) Personalized Disease Networks in Predicting Cardiovascular Outcomes
Generating Network Matrix
Description
This function use data set with cut off information to create network matrix
Usage
buildnetworks(x, k1, del = 0)
Arguments
x |
comorbidity data matrix with column correspond to ICD9 codes and row correspond to each patient |
k1 |
the cut off point between Diagnosis/Procedure A to another Diagnosis/Procedure B, it can be fix number, NULL and datecut |
del |
number of character deleted for each name of the input |
Value
Network Matrix
Examples
# Select a subset of data for toy example
comorbidity_data = comorbidity_data[c(1:10),]
survival_data = survival_data[c(1:10),]
k1 = datecut(comorbidity_data,survival_data[,1],survival_data[,2])
a = buildnetworks(comorbidity_data,k1)
Sampele comorbidity data set
Description
This is the sample data set for Package PDN
Usage
comorbidity_data
Format
100 rows and 10 column matrix
Generating Optimal cuts for the Data
Description
Performs Cox proportional hazards regression model on patients survival days based on different cutoffs
Usage
datecut(x, surdays, event)
Arguments
x |
comorbidity data matrix with column correspond to ICD9 codes and row correspond to each patient |
surdays |
survival days for each patient |
event |
indictor variable 1 represent patient died 0 represent patient survive |
Details
Building and draw personized disease network
Value
a vector of cutoff points that maximize the Z statistics for cox model between each Diagnosis/Procedure A to another Diagnosis/Procedure B
Examples
comorbidity_data
survival_data
#Select a subset of data for toy example
comorbidity_data = comorbidity_data[c(1:10),]
survival_data = survival_data[c(1:10),]
k1 = datecut(comorbidity_data,survival_data[,1],survival_data[,2])
Sampele demographic data set
Description
This is one of the sample data set. It contains the demographic information of each patients It has five variables which are sex, race, hispan, dshyr and prime
Usage
demographic_data
Format
100 rows and 5 column matrix
Draw Personalized Disease Network for one patient with network and ggplot2
Description
Draw Personalized Disease Network based on newtwork matrix
Usage
draw.PDN(tt, labels)
Arguments
tt |
one row of network matrix generated from comorbidity data matrix using buildnetworks |
labels |
names of each node in the network matrix |
Examples
#Select a subset of data for toy example
comorbidity_data = comorbidity_data[c(1:10),]
survival_data = survival_data[c(1:10),]
# Getting the names
k1 = datecut(comorbidity_data,survival_data[,1],survival_data[,2])
a = buildnetworks(comorbidity_data,k1)
#Plot PDN for patient 7
nn = names(comorbidity_data)
draw.PDN(a[7,],nn)
Draw Personalized Disease Network for one patient
Description
Draw Personalized Disease Network based on newtwork matrix
Usage
draw.PDN.circle(a0, dak)
Arguments
a0 |
one row of network matrix generated from comorbidity data matrix using buildnetworks |
dak |
one row of Ranks matrix for corresponding comorbidity data matrix |
Examples
#Select a subset of data for toy example
comorbidity_data = comorbidity_data[c(1:10),]
survival_data = survival_data[c(1:10),]
# Find date cuts
k1 = datecut(comorbidity_data,survival_data[,1],survival_data[,2])
# Build networks
a = buildnetworks(comorbidity_data,k1)
# Graph individual patients
datark = t(apply(comorbidity_data,1,rank))
dak = sort(datark[1,])
# draw PDN for the first patient
draw.PDN.circle(a[1,],dak)
# draw PDN for the whole comorbidity data set
par(mfrow=c(2,5))
for(i in 1 : nrow(a)){
dak = apply(datark,2,sort)
draw.PDN.circle(a[i,],dak[i,])
title(main=paste("Patient",i))
}
Draw Personalized Disease Network for cluster of patients
Description
Draw Personalized Disease Network based on cluster information
Usage
draw.PDN.cluster(a0, dak)
Arguments
a0 |
network matrix get from buildnetworks |
dak |
ranks data for comorbidity data matrix |
Examples
#Select a subset of data for toy example
comorbidity_data = comorbidity_data[c(1:10),]
survival_data = survival_data[c(1:10),]
##Clustering Example
k1 = datecut(comorbidity_data,survival_data[,1],survival_data[,2])
a = buildnetworks(comorbidity_data,k1)
datark = t(apply(comorbidity_data,1,rank))
require(survival)
zsq = NULL
for(i in 1:ncol(a)){
a1 = (summary(coxph(Surv(as.numeric(survival_data[,1]),survival_data[,2]) ~ a[,i],
data=as.data.frame(a)))$coefficient[,4])^2
zsq = cbind(zsq,a1)
}
zsq = as.numeric(zsq)
wi=zsq/sum(zsq,na.rm=TRUE)
wi[wi<10^ -3]=10^ -3
wi[is.na(wi)]=10^ -3
#weighted matrix
wa = NULL
for(i in 1:ncol(a)){
wa = cbind(wa,a[,i]*wi[i])
}
#PCA
pr.out=prcomp(wa)
x.svd=svd(wa)
x.score1 <- wa %*% x.svd$v
x.score2 <- x.svd$u %*% diag(x.svd$d)
##HC cluster using the first 8 PCA scores
dp<-dist(x.score2[,1:8])
hcp<-hclust(dp, method="ward.D")
##Plot Network
s1=rev(sort(apply(a[cutree(hcp,3)==2,],2,mean)))[1:50]
dak = sort(apply(datark[cutree(hcp,3)==2,],2,mean,na.rm=TRUE))
names(dak) = unlist(strsplit(names(dak),"DAT"))
draw.PDN.cluster(s1,dak)
Sampele Survival data set
Description
This is the sample data set for function datecut. The first column is the time difference between the patient admission date and death date The second column is the indictor variable with 1 corresponding to patient death, 0 corresponding to patient alive
Usage
survival_data
Format
100 rows and 2 column matrix