In neellab/simem: Prediction of differential gene essentiality from pooled phenotypic screens (eg: siRNA, shRNA, CRISPR/Cas9 screens) using linear mixed effect models.

Formatting Screen Data for Analysis

 

Pooled screen measurements are highly structured. Each measurement is associated with a specific gene, reagent (siRNA pool, shRNA, gRNA), cell line, replicate and/or time-point. To manage the measurements and extensive annotations, we bundle them all in a Bioconductor ExpressionSet object. Before detailing the formatting process, here are links to pre-formatted and normalized ExpressionSets for several published, large-scale pooled screens:

• Marcotte et al. 2016's 77 breast screens ExpressionSet RData file
• Cowley et al. 2014's 216 multi-cancer screens ExpressionSet RData file. Also see the extensive resources at the Broad Institute's Project Achilles website.
• Cheung et al. 2011's 102 multi-cancer screens ExpressionSet RData file. Also see the extensive resources at the Broad Institute's Project Achilles website.

 

Here's what that looks like for the breast screens:

# To install Bioconductor packages Biobase and preprocessCore, uncomment:
# source("http://www.bioconductor.org/biocLite.R")
# biocLite()

suppressPackageStartupMessages(library(Biobase))
library(preprocessCore)
library(ggplot2)

measurements = read.delim("../../data/breast_measurements.txt", header=T, as.is=T, check.names=F)
reagents = read.delim("../../data/breast_reagents.txt", header=T, as.is=T, check.names=F)
samples = read.delim("../../data/breast_samples.txt", header=T, as.is=T, check.names=F)

# Sanity check that the sample names match between measurements matrix and sample information data frame
table(colnames(measurements) == samples$sample_id, useNA="always")
rownames(samples) = samples$sample_id


# It's important to ensure that the measurements matrix and reagents information data frame have the same reagent ordering (reagents$trcn_id in this case)

rownames(reagents) = reagents$trcn_id
rownames(measurements) = reagents$trcn_id

reagentsADF = new("AnnotatedDataFrame", reagents)
samplesADF = new("AnnotatedDataFrame", samples)

# Create a mew ExpressionSet combining and linking measurements, reagent information and sample information
screens = new("ExpressionSet",
              exprs=measurements,
              featureData=reagentsADF,
              phenoData=samplesADF)

dir.create("../../data/results")

# Save the ExpressionSet as an RData file, suffixed with ".eset" to clarify that it contains an ExpressionSet
# This file can subsequently be loaded into R using the load() function
save(screens, "../../data/results/breast_screens.eset")

neellab/simem documentation built on May 23, 2019, 1:31 p.m.