Drug response prediction

Oracle VM VirtualBox
we provide a VirtualBox image file. It's set up ready to run this program with testing data set.

How to set up:
Instructions for Mac or Linux
Install Virtual Box and Import the Virtual Machine on Mac or Linux
1. Install VirtualBox
a. Select the VirtualBox installer for Mac
b. Click through the dialogs and install VirtualBox
c. Restart your computer
d. Find "Oracle VM VirtualBox" in your applications and open the program
2. In the VirtualBox window, select File -> Import Appliance...
3. Click Open Appliance...
4. Select the Ubuntu-64-DR.ova
5. Click Next
6. Ensure that the Reinitialize the MAC address of all selected network cards option is selected
7. Click Import, this will start importing the VM and the process may take several minutes
8. Verify that the Virtual Machine was successfully imported by selecting the new VM from the list in Virtual Box and clicking Start. A new window will open and begin loading your VM, if you are prompted to enter a password you have completed setting up your local dev environment.

SSH into your local Virtual Machine
1. On your local machine, open a terminal by launching the Terminal application
2. In your terminal's command line, please enter the following: ssh vannberglab@localhost -p10022 and hit Enter
3. You will be prompted for a password, the password is vannberglab

Instructions for Windows Users
Install Virtual Box and Import the Virtual Machine on Windows
1. Install VirtualBox
a. Select the appropriate VirtualBox installer for your OS
b. Click through the dialogs and install VirtualBox
c. Restart your computer
d. Find "Oracle VM VirtualBox" in your applications and open the program
2. In the VirtualBox window, select File -> Import Appliance...
3. Click Open Appliance...
4. Select the Ubuntu-64-DR.ova
5. Click Next
6. Ensure that the Reinitialize the MAC address of all selected network cards option is selected
7. Click Import, this will start importing the VM and the process may take several minutes
8. Verify that the Virtual Machine was successfully imported by selecting the new VM from the list in Virtual Box and clicking Start. A new window will open and begin loading your VM, if you are prompted to enter a password you have completed setting up your local dev environment.

SSH into your local Virtual Machine
1. Download and install an SSH client if you do not already have one. Some examples of popular SSH
clients: PuTTY,OpenSSH, and WinSCP
2. Open your SSH client, the following example will be using PuTTY
3. Fill in the following information:
Host Name (or IP Address): vannberglab@localhost Port: 10022
4. Click Open
5. A terminal will open and you will be prompted for a password, the password is vannberglab


Figures

1. bash scripts 
	figures_run.sh
	run for all figures
	
2. 1 3 model compare bar and roc  
		R scripts:
		source("bar.r")
		source("roc.r")

3. F2c GEO datasets density plot
	All 3 GEO datasets are pre-processed with pipeline script from KEA Prediction_R package. Detailed readme file is provided there.
	R script:
	source("density_plot.r")

4. F2d Response rate regression
	scatter plot of predicted and observed ressponse rate and linear regression



R package use for predicting drug response of patients


Predicting drug response of patients

This function predict drug response by using svm based recersive feature elimination.


Requirement

library affy
install instruction:
source("http://bioconductor.org/biocLite.R")
biocLite("affy")

Usage

DrugResponse.predict(input file, inputfile_formate, sample_name, path_to_model_files’../data')

Arguments

patient_inputfile		gene expression data for patient.

inputfile_formate     		input file formate, can be cel (cel file) or exp (gene expression file)

patient_name     		patient name

path.to.model     		where the model stored, default to '../data_fix'. 
 
return a matrix of positive or negative score for each drug

output pdf files of prediction score


Install

Download zip file and use install.package function in R to install
R CMD INSTALL --merge-multiarch DrugResponse_*.tar.gz
Unzip file and prepared model files in DrugResponse/data_fix folder (this model is pre-build by DrugResponse model program), and test files in DrugResponse/test_data folder


Example

input exp file
DrugResponse.predict('../data/test_data/545R_992om_3ex.tsv','exp',cbind('545R','992om'),'../data_fix')

input cel file
DrugResponse.predict(patient_inputfile,'cel',patient name,'../data_fix')

RMD bacth
R -q -e "library(DrugResponse); library(affy); DrugResponse.predict('../data/test_data/545R_992om_3ex.tsv','exp',cbind('545R','992om'),'../data_fix')"


Pipeline

we provide a bash script to run test for cancer samples in pipeline folder
DrugResponse_sample_test_GSE.sh 
default model files is ../data_fix  You may change it to where you unzip the DrugResponse pacakge path_to_DrugResponse/data_fix.
input: GSE number

example (path here is for virtual box file system)

run test for GSE sets
# ./home/vannberglab/DrugResponse/pipeline/DrugResponse_sample_test_GSE.sh GSE30161

run test for GSM samples
# ./home/vannberglab/DrugResponse/pipeline/DrugResponse_sample_test_GSE.sh GSM1167083

run test for cancer samples (CEL files)
#./home/vannberglab/DrugResponse/pipeline/DrugResponse_sample_test_URL.sh 'path.to.test.samples.folder'

run normalization for GSE sets
# ./home/vannberglab/DrugResponse/pipeline/matrix_normalization/DrugResponse_sample_norm_GSE.sh GSE30161