CG-TARGET

CG-TARGET is a collection of R scripts to be used for the purpose of predicting gene-set targets from chemical-genetic interaction profiles.

Features

• Interprets chemical-genetic interaction profiles using a reference set of genetic interaction networks.
• Species-independent: any species with chemical-genetic interaction profiles can be intepreted, as long as an appropriate genetic interaction dataset is available.
• Genetic interaction datasets and gene set annotations are automatically downloaded. (users can supply their own GI datasets and gene sets as well)
• Scalable: amenable to the analysis of both small and large chemical-genetic interaction datasets.
• Written in R; no other dependencies.

How to use

After downloading the software and setting up your environment variables, it is quite simple to run the commands. First, you will need to set up your configuration file (here called config_file.yaml). The example_config_file.yaml is provided as a template in the main folder.

General overview

Given the following datasets, CG-TARGET predicts the biological processes (or other functionally coherent gene sets) that are perturbed by compounds:

1. A dataset of chemical-genetic interaction profiles obtained for compounds and negative controls (typically a solvent control such as DMSO) against a mutant collection
2. A dataset of genetic interaction profiles obtained for "query" gene mutants against a mutant collection that overlaps with the profiles in (1)
3. A collection of biological process (or other gene set) annotations for the query genes in (2)

CG-TARGET can be broken down into four distinct steps:

1. gen_randomized_profiles.r generates randomly-resampled profiles, which are used in z-score/p-value and FDR calculations, from compound-treated conditions.
2. predict_gene_targets.r integrates the chemical-genetic and genetic interaction profiles to generate a similarity score for each combination of compound and genetic interaction query gene.
3. predict_gene_set_targets.r aggregates these compound-gene similarity scores at the level of biological processes (or other gene sets) to generate z-scores and p-values for each compound-process association.
4. gene_set_FDR_analysis.r uses the process-level predictions derived from negative control and resampled chemical-genetic interaction profiles to calculate estimates of the false discovery rate.

Script details

Note: config_file.yaml defines a folder that contains the output of all scripts, called output_folder. This is referenced below as <output_folder>.

View available genetic interaction datasets

gi_datasets.r

Use this script to check which genetic interaction datasets are available.

View available gene sets

gene_sets.r

Use this script to check which gene set annotations are available.

Generate resampled profiles

gen_randomized_profiles.r config_file.yaml

This script generates the set of resampled profiles derived from compound-treated conditions. Output is in <output_folder>/resampled_profiles/.

Predict gene targets for both the real and resampled profiles

predict_gene_targets.r config_file.yaml
predict_gene_targets.r --rand config_file.yaml

This script generates compound-gene similarity scores by integrating chemical-genetic and genetic interaction profiles. Output is in <output_folder>/gene_target_prediction/. Without the --rand flag, the computations are performed for the compound-treated and negative control profiles, and with the --rand flag, they are performed for the resampled profiles.

Optional: Create a clustered heat map of gene target prediction scores

This will be a *.CDT file, viewable with Java TreeView. You can choose to view the predictions for the resampled profiles too (again, using --rand).

visualize_gene_targets.r config_file.yaml
visualize_gene_targets.r --rand config_file.yaml

Predict the gene-set targets

predict_gene-set_targets.r config_file.yaml

This script aggregates compound-gene similarity scores into z-scores and p-values for each combination of compound and biological process. Output is in <output_folder>/gene_set_target_prediction/.

Estimate false discovery rate and export final gene-set target prediction tables.

gene_set_FDR_analysis.r config_file.yaml

This script compares the rate of prediction for compounds versus 1) negative control profiles and 2) resampled profiles across the full range of p-values, resulting in empirical estimates of the false discovery rate. Output is in <output_folder>/final_results/.

Installation

Requirements

This software is written in R, and thus requires a working R installation. Microsoft Open R (or any other R with an optimized BLAS library, such as OpenBLAS) is recommended, as speed at which some steps finish depends highly on the speed of the matrix multiplications involved.

The following libraries are required:

data.table
digest
ggplot2
grid
gridExtra
optparse
reshape2
tools
yaml

The following system commands are required:

gzip (the function of `gzip -dc` must be to decompress a gzipped file and print to stdout)

The following libraries are optional:

(they are only required for the visualize_gene_targets.r script, which is optional)

ctc
fastcluster

Downloading CG-TARGET

Basic

Head on over to https://github.com/csbio/CG-TARGET/releases/ and download the latest release. Extract the compressed folder to a good location from which to run the software (i.e. get it out of your downloads folder!)

Advanced

If you know what you are doing and want to keep up-to-date with the latest version, clone the repository (git clone https://github.com/csbio/CG-TARGET.git or windows equivalent).

Setting up environment variables

Required: TARGET_PATH

Set the value of this environment variable to the path of the CG-TARGET folder you downloaded and extracted. The scripts from BEAN-counter will look for this variable in your environment, so it must be set!

Optional, but strongly recommended: adding $TARGET_PATH/scripts/ to your PATH

Adding the scripts folder inside of the CG-TARGET folder to your PATH environment variable allows you to execute the commands in the pipeline by calling them only by their names (all demos will assume this has been done).

How do I set my environment variables?

Linux/Mac

The best way to do this is by adding code to the scripts that run every time you open a new shell. If you use the bash shell, then add the following line to either your ~/.bashrc or ~/.bash_profile files:

export TARGET_PATH=/your/path/to/CG-TARGET/

If you use the c shell (csh), then add the following line to your ~/.cshrc file:

setenv TARGET_PATH /your/path/to/CG-TARGET/

(but please, please do us all a favor and ditch the c shell already)

To append a directory to your PATH variable, add this line to your ~/.bashrc or ~/.bash_profile (or equivalent for ~/.cshrc):

export PATH=$PATH:/your/path/to/CG-TARGET/scripts

Windows

Tutorial for changing path variable

License

Use of this code is free for academic use and for a 60-day commercial trial period. Sustained commercial use requires the purchase of a license.

In any case, please head on over to z.umn.edu/cgtarget to register for a license. This helps me keep track of the user base :)