CeRberus

The CeRberus module is part of the Cerberus package and performs the final steps of aggregating guide-level genetic interaction (GI) scores to gene-pair level scores. It can also be used independently as part of other R GI scoring pilelines.

Quick start

1 Install devtools from CRAN:

if (!require("devtools", quietly = TRUE)) {
	install.packages("devtools")
	}

2. Install the CeRberus package:

devtools::install_github("thorohde/CeRberus")

3. Follow the Bioconductor installation routine to install the limma package:

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("limma")

4. Load the CeRberus package into the workspace:

library(CeRberus)

5. After storing guide GI scores in a csv file, call the modules full_run() routine with an instruction yaml file:

full_run(yaml_fpath = "input_yaml_file.yaml")

Input parameters

parameter	description	required
scores_file	path to a csv file of guide GI scores	*
output_directory	path where the package stores the computed scores	*
FDR	method used for to estimate the False Discovery rate. Default: BH
skip_update	Can be used to suppress package updates.
verbose	should the pipeline print a feedback while running?
keep_all_configurations	Skip discarding configurations with negative duplicateCorrelation.

Input file format

The input table of guide GI scores is required to be in long format, one observation per row. The columns bio_rep, tech_rep and guide_pair describe the experimental setup and will be used as replication layers by CeRberus.

bio_rep	tech_rep	guide_pair	query_gene	library_gene	GI
b1	t1	g1	IGF2	ADRB2	1.02869961
b1	t1	g2	IGF2	ADRB2	-0.46078453
b1	t1	g3	IGF2	ADRB2	1.01662498
b1	t2	g1	IGF2	ADRB2	-0.94890821
b1	t2	g2	IGF2	ADRB2	-0.44307736
b1	t2	g3	IGF2	ADRB2	-0.5146814
b2	t1	g1	MTOR	IMPDH1	-1.1482365
b2	t1	g2	MTOR	IMPDH1	-2.6873593
b2	t1	g3	MTOR	IMPDH1	-0.6714192
b2	t2	g1	MTOR	IMPDH1	-0.79467931
b2	t2	g2	MTOR	IMPDH1	-2.05650114
b2	t2	g3	MTOR	IMPDH1	-0.32426398

Procedure

1. After importing the guide GI scores, CeRberus infers the experimental structure based on the query and library gene set.

• Fixed Pair Screen: Query genes and library genes are different.
• Multiplex Screen: Query and library gene sets are identical.

2. CeRberus tests which configuration produces desirable duplicateCorrelation values. A positive correlation between either the biological or technical replicates is expected.

3. For the configurations that pass these filters, genetic interactions are computed, using linear model fits. P-values are computed using Empirical Bayes Statistics, and corrected for multiple testing by a method of choice.

4. The computed scores are exported to the given output directory.