DEGEF

Differentially Expressed Gene Expression Finder (DEGEF) discovers genomic regions of enriched differential gene expression from differential gene expression results.

Installation

We recommend running DEGEF using Python-3.7 or greater. Required python packages include argparse, numpy, pandas, random, scipy, statsmodels, and matplotlib.

To install DEGEF, run the following code:

git clone --recursive https://github.com/gaog94/DEGEF

Then add the pulled directory to PYTHONPATH.

export PYTHONPATH="/path/to/DEGEF/directory/:$PYTHONPATH"

DEGEF is able to run using differential gene expression results in table form as input. The differential expression table should include the following columns: xyz, xyz, xyz.

To generate these results, run a differential gene expression finder like DESeq2 or limma-voom.

featureCounts is available as part of the Rsubread/Subread package, which may be obtained at https://bioconductor.org/packages/release/bioc/html/Rsubread.html.

limma-voom may be obtained at https://bioconductor.org/packages/release/bioc/html/limma.html.

Description

DEGEF (rhymes with "VEGF") stands for Differentially Expressed Gene Expression Finder and is an algorithm that searches for genomic areas of enriched differential gene expression. Using differential gene expression results from common differential expression analysis pipelines such as limma-voom or DESeq2, DEGEF scans the genome searching for regions with greater differential expression than expected by chance.

Details of implementation are included in our manuscript "Co-localization of Clusters of TCR-Regulated Differentially Expressed Genes with Areas of TAD Rearrangement" (in preparation).

Usage

Running DEGEF in Python

After adding the pulled directory to PYTHONPATH, run the following:

import DEGEF

peakGenesTot, peakSigsTot = DEGEF.run_enrichment_finder(filename, fnRef, fnSize, \
	                                direction=direction, metric=metric, p=p,\
	                                logFC=logFC, maxWindow=maxGeneInWindow, \
	                                nboot=nboot,windowSize=windowSize, \
	                                stepSize=stepSize, writeToDisk=True, \
	                                outDirName=outDirName)

Running DEGEF from the Command Line

After generating a differential gene expression results table using differential expression analysis software (e.g., limma-voom, DESeq2, etc.), run the following:

python DEGEF.py -i filename -o outDir -l refLoci -c refSizes -m metric -d dir -w ws -s ss -p pval -f lfc -x max -n nBoot -q fdr

The following input parameters are used:

• -i --inputDEGFile filename is the input tsv file outputted by a differential expression analysis pipeline.
• -o --outputDirectory outDir is the output directory where DEGEF will write its output files.
• -l --refTranscriptLoci refLoci is the path to the reference file containing transcripts' genomic loci. (Default: "../ReferenceFiles/hg38_ensemble_loci_modified.tsv")
• -c --refChromSizes refSizes is the path to the reference file containing chromosome sizes. (Default: "../ReferenceFiles/chrom_hg38.sizes")
• -m --scoringMetric metric is the scoring metric ("count" vs "significance" vs "foldchange") to use for calculating raw scores. (Default: "foldchange")
• -d --scoringDirection dir is the directionality of expression ("up" vs "dn" vs "mx") that DEGEF will test for enrichment in. (Default: "up")
• -w --windowSize ws is the size of the genomic window used to compute enrichment scores. (Default: 1e6)
• -s --stepSize ss is the number of basepairs the genomic window shifts by when computing enrichment scores. (Default: 2e5)
• -p --limmaPValueThreshold pval is the adjusted p-value threshold used to classify genes as up/down/not regulated when computing raw scores. (Default: 0.05)
• -f --limmaLogFoldChangeThreshold lfc is the log-fold-change threshold used to classify genes as up/down/not regulated when computing raw scores. (Default: 0.58)
• -x --maxGeneInWindow max is the maximum number of genes within a window before DEGEF will switch from bootstrapping to using the Central Limit Theorem (CLT) to compute the null distribution of enrichment scores. (Default: 30)
• -n --nBootstraps nBoot is an integer representing the number of bootstraps to conduct when building null distributions for the enrichment score. (Default: 6400)
• -q --fdrPeakThreshold fdr represents the False Discovery Rate (FDR) threshold to use to define DEGEF enrichment peaks. (Default: 0.05)

Attributions

KIRCLE is developed and maintained by Galen Gao -- Laboratory of Molecular Immunology -- National Heart Lung and Blood Institute (NHLBI), National Institutes of Health (NIH), Bethesda, MD.