Submitting to DREAM challange

Document

Main information can be foud in this page: https://www.synapse.org/#!Synapse:syn20940518/wiki/600165

Step 1: Create Rscript to create the model

Example of a model created:

model.list <- readRDS("models/modelsSC1_rna_seq_only.rds")
class(model.list)

## [1] "list"

names(model.list)

##   [1] "17-AAG (Tanespimycin)"      "A-674563"                  
##   [3] "ABT-737"                    "Afatinib (BIBW-2992)"      
##   [5] "Alisertib (MLN8237)"        "AT7519"                    
##   [7] "Axitinib (AG-013736)"       "AZD1480"                   
##   [9] "Barasertib (AZD1152-HQPA)"  "Bay 11-7085"               
##  [11] "BEZ235"                     "BI-2536"                   
##  [13] "BMS-345541"                 "Bortezomib (Velcade)"      
##  [15] "Bosutinib (SKI-606)"        "Cabozantinib"              
##  [17] "Canertinib (CI-1033)"       "Cediranib (AZD2171)"       
##  [19] "CHIR-99021"                 "CI-1040 (PD184352)"        
##  [21] "Crenolanib"                 "Crizotinib (PF-2341066)"   
##  [23] "CYT387"                     "Dasatinib"                 
##  [25] "DBZ"                        "Doramapimod (BIRB 796)"    
##  [27] "Dovitinib (CHIR-258)"       "Elesclomol"                
##  [29] "Entospletinib (GS-9973)"    "Entrectinib"               
##  [31] "Erlotinib"                  "Flavopiridol"              
##  [33] "Foretinib (XL880)"          "GDC-0879"                  
##  [35] "GDC-0941"                   "Gefitinib"                 
##  [37] "Gilteritinib (ASP-2215)"    "Go6976"                    
##  [39] "GSK-1838705A"               "GSK-1904529A"              
##  [41] "GSK690693"                  "GW-2580"                   
##  [43] "Ibrutinib (PCI-32765)"      "Idelalisib"                
##  [45] "Imatinib"                   "INK-128"                   
##  [47] "JAK Inhibitor I"            "JNJ-28312141"              
##  [49] "JNJ-38877605"               "JNJ-7706621"               
##  [51] "JQ1"                        "KI20227"                   
##  [53] "KU-55933"                   "KW-2449"                   
##  [55] "Lapatinib"                  "Lenalidomide"              
##  [57] "Lenvatinib"                 "Lestaurtinib (CEP-701)"    
##  [59] "Linifanib (ABT-869)"        "Lovastatin"                
##  [61] "LY-333531"                  "Masitinib (AB-1010)"       
##  [63] "MGCD-265"                   "Midostaurin"               
##  [65] "MK-2206"                    "MLN120B"                   
##  [67] "MLN8054"                    "Motesanib (AMG-706)"       
##  [69] "Neratinib (HKI-272)"        "NF-kB Activation Inhibitor"
##  [71] "Nilotinib"                  "Nutlin 3a"                 
##  [73] "NVP-ADW742"                 "NVP-TAE684"                
##  [75] "Palbociclib"                "Panobinostat"              
##  [77] "Pazopanib (GW786034)"       "PD173955"                  
##  [79] "Pelitinib (EKB-569)"        "PHA-665752"                
##  [81] "PHT-427"                    "PI-103"                    
##  [83] "PLX-4720"                   "Ponatinib (AP24534)"       
##  [85] "PP242"                      "PRT062607"                 
##  [87] "Quizartinib (AC220)"        "RAF265 (CHIR-265)"         
##  [89] "Rapamycin"                  "Regorafenib (BAY 73-4506)" 
##  [91] "Roscovitine (CYC-202)"      "Ruxolitinib (INCB018424)"  
##  [93] "S31-201"                    "Saracatinib (AZD0530)"     
##  [95] "SB-431542"                  "Selinexor"                 
##  [97] "Selumetinib (AZD6244)"      "SGX-523"                   
##  [99] "SNS-032 (BMS-387032)"       "Sorafenib"                 
## [101] "SR9011"                     "Staurosporine"             
## [103] "STO609"                     "SU11274"                   
## [105] "Sunitinib"                  "Tandutinib (MLN518)"       
## [107] "TG100-115"                  "TG101348"                  
## [109] "Tivozanib (AV-951)"         "Tofacitinib (CP-690550)"   
## [111] "Tozasertib (VX-680)"        "Trametinib (GSK1120212)"   
## [113] "Vandetanib (ZD6474)"        "Vargetef"                  
## [115] "Vatalanib (PTK787)"         "Vemurafenib (PLX-4032)"    
## [117] "Venetoclax"                 "Vismodegib (GDC-0449)"     
## [119] "Volasertib (BI-6727)"       "VX-745"                    
## [121] "XAV-939"                    "YM-155"

model.list[[1]]

## Random Forest 
## 
##  184 samples
## 1000 predictors
## 
## No pre-processing
## Resampling: Cross-Validated (10 fold, repeated 3 times) 
## Summary of sample sizes: 165, 165, 164, 165, 166, 167, ... 
## Resampling results:
## 
##   RMSE      Rsquared   MAE     
##   38.41165  0.3336447  31.06616
## 
## Tuning parameter 'mtry' was held constant at a value of 31.63858

Step 2: Create Rscript to make predictions using the model

Subchallenge 1 (Source: https://www.synapse.org/#!Synapse:syn20940518/wiki/600161)

Predict ex-vivo drug sensitivity using clinical and genomic features of tumors. Drug sensitivity is measured as the area under the dose response curve (AUC).

Assessment Each drug will be evaluated using a Spearman correlation([0, 1]), and the Spearman correlations will be averaged to produce a final metric. (Davidson-Pilon et al. 2019)

Output The Docker container must write a csv at /output/predictions.csv with one row per (specimen, inhibitor) pair and three columns:

• lab_id: specimen identifier
• inhibitor: name of the drug
• auc: predicted AUC

The list of all inhibitors can be pulled from aucs.csv.

run_model_SC1.R code

#------------------------------------------------
# Dream challange SC1 prediction Rscript  
# - Drescription: 
#    Reads input files in /input/ folder
#    Use models to predict drug score
#    Output prediction scores in /output/ folder
# - Author: Tiago C. Silva
# - Date: 02/12/2020
#------------------------------------------------

#--------------------------------------
# Loading libraries
#--------------------------------------
library(readr, quietly = TRUE)
library(plyr, quietly = TRUE)
library(dplyr, quietly = TRUE)
library(caret, quietly = TRUE)
library(matrixStats, quietly = TRUE)

#--------------------------------------
# INPUT
#--------------------------------------
# Reading the data
input.dir <- "/input/"
input.files <- dir(input.dir,full.names = TRUE)
input.dfs <- lapply(input.files, function(f) read_csv(f,col_types = readr::cols()))
names(input.dfs) <- gsub(".csv","",basename(input.files))

# Treating input the same way input training data
rna <- input.dfs$rnaseq[,-c(1:2)] %>% t %>% as.data.frame() 
colnames(rna) <- input.dfs$rnaseq$Symbol
rna.var.order.idx <- rna %>% as.matrix %>% colVars %>% order(decreasing = TRUE)
rna.zscore <- apply(rna, 2,scale) %>% as.data.frame()
rownames(rna.zscore) <- rownames(rna)

#--------------------------------------
# MODEL
#--------------------------------------
model <- readRDS("/usr/local/bin/modelsSC1_rna_seq_only.rds")

# Prediction
output_df <- plyr::adply(names(model),1,function(drug){
    variables <- names(model[[1]]$trainingData)[-1]
    auc <- predict(model[[drug]], rna.zscore) 
    data.frame("lab_id" = names(auc),
               "inhibitor" = rep(drug,length(auc)), 
               "auc" = auc)
},.progress = "time",.inform = TRUE,.id = NULL)

#--------------------------------------
# Output
#--------------------------------------
output.dir <- "/output/"
write_csv(output_df, file.path(output.dir,"predictions.csv"))

Step 3: Create the docker image

Your docker file will create the enviroment you need to run the prediction model. You will need to install the required libraries, add the models into it, and create an Rscript to predict and output the data.

Observation: FROM r-base will provide R 3.6.2.

Dockerfile code

## Start from this Docker image
FROM r-base

## Install R in Docker image
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y r-base

## Install R packages in Docker image
RUN Rscript -e "install.packages('BiocManager')"
RUN echo "r <- getOption('repos'); r['CRAN'] <- 'http://cran.us.r-project.org'; options(repos = r);" > ~/.Rprofile
RUN Rscript -e "install.packages('readr')"
RUN Rscript -e "install.packages('caret')"
RUN Rscript -e "install.packages('matrixStats')"
RUN Rscript -e "install.packages('ranger')"
RUN Rscript -e "install.packages('randomForest')"

## Copy your files into Docker image
COPY run_model_SC1.R /usr/local/bin/
COPY models/modelsSC1_rna_seq_only.rds /usr/local/bin/
RUN chmod a+x /usr/local/bin/run_model_SC1.R

## Make Docker container executable
ENTRYPOINT ["Rscript", "/usr/local/bin/run_model_SC1.R"]

Step 4: build, test and send to synapse

The script below is an example to make it easier to:

• build the docker with command: ./submit.sh build
• test the docker with command: ./submit.sh test
• send the docker to snapse with command: ./submit.sh send

The SYNAPSE_PROJECT_ID should be modified by each student. To make the bash script executable run chmod +x submit.sh

submit.sh code

#!/bin/bash

SYNAPSE_PROJECT_ID="syn21557290"
echo $SYNAPSE_PROJECT_ID
if [ "$1" == "test" ]; then
    echo "Testing docker"
    docker run -v "$PWD/input/:/input/" -v "$PWD/output:/output/" docker.synapse.org/$SYNAPSE_PROJECT_ID/sc1_model 
elif [ "$1" == "build" ]
then
    echo "Building docker ...."
    docker build -t docker.synapse.org/$SYNAPSE_PROJECT_ID/sc1_model .
elif [ "$1" == "send" ]
then
    echo "Sending docker ...."
    docker login docker.synapse.org
    docker push docker.synapse.org/$SYNAPSE_PROJECT_ID/sc1_model
else
    echo "Options available are: test or send"
fi

Step 5: submit docker image in synapse

Select the docker submitted and click on Submit Docker Repository to Challenge.

Step 6: Check results

You will receive an email with the success or failure (with logs).

You should be able to check you results at: https://www.synapse.org/#!Synapse:syn20940518/wiki/600158

score