MEAGA

Introduction

Pathway analysis for results from genetic association studies could help us better understand complex traits. MEAGA (Minimum distance-based Enrichment Analysis for Genetic Association) performs functional/pathway enrichment test while integrating network information from biological interactome (e.g. protein-protein interaction network) using graphical algorithm techniques.

The latest version of MEAGA could be obtained in here:

• MEAGA_1.2.tar.gz

How it works

MEAGA tests the hypothesis that genes from the susceptibility loci in the trait/disease-associated function/pathway are closer with each other in the biological interactome. MEAGA takes the markers used in the association results as input. Users would pre-specify the association signals and annotate the tagged genes for each marker (e.g. using linkage disequilibrium- or genomic distance- based block).

For each functional gene-set being tested, MEAGA first identifies the overlapping genes from the signals, then utilizes graphical algorithms (Kou's algorithm to identify Steiner Tree(s)) to construct subgraph(s) with minimum distance(s) in the interactome. MEAGA computes a statistic (S) summarizing the amount of overlapping genes and the overall shortest distance(s) of the subgraph(s). MEAGA uses sampling strategy to approximate the null distribution of S and compute empirical and multiple testing-corrected p-values.

MEAGA was implemented in Python, and it requires the graphical features from NetworkX (Hagberg et al. 2008): https://networkx.github.io/.

The overview workflow of MEAGA:

Usage instructions

Input Files

Functional/Pathway annotation file

Three-column file annotates the gene's associated functions/pathways (2nd column) and the source (e.g. Gene Ontology)

We provided a pre-compiled functional annotation file:

./db/gene2fun.txt

We obtained the functional and pathway annotation data from the GO (Ashburner et al. 2000), KEGG (Kanehisa et al. 2012), and Reactome (Croft et al. 2013) databases. We processed the GO’s gene-to-GO file so we also annotated each gene with the “ancestral” terms of its annotated term(s).

Marker to Gene annotation file

A 5-column file for marker to gene annotation. The first four columns are the same as plink-map file, and the last column indicates the annotated genes (separated by semicolon).

Example:

./example/marker2gene.txt

Associated signals

A 1-column file for the best signals identified in the genetic association study

Example:

./example/intmarkers

Command References

Option	Name	Descritption	Required	Default value
-s	--marker2gene	marker to gene annotation file	Yes	NA
-g	--gene2fun	function/pathway annotation file	Yes	NA
-i	--intmarkers	associated signals	Yes	NA
-d	--funSPdir	directory storing the prefix folders for each function shortest-path files	Yes	NA
-o	--outDir	output directory and prefix	Yes	NA
-t	--numProcess	number of process used	No	1
-n	--numsamples	number of samplings used to construct null distribution of S	No	10,000
-m	--minFun	only used function/pathway with this minimum number of genes	No	5
-M	--maxFun	only used function/pathway with this maximum number of genes	No	1000
-a	--minIntFun	minimum number of associated genes overlapped with the genes in the function/pathway	No	3
-D	--Dmax	distance value set for genes not connected in the interactome	No	12

If you need help in understanding the options, you could type:

./bin/MEAGA.py --help

Output files

There are two final output files for MEAGA: MEAGAresult and MEAGAtree.

Descriptions for MEAGAresult columns

Funs	NumFunGenes	NumFunIntGenes	NumConnectedGraphs	AvgIntFunGenes_perConnectedGraphs	S	pval	adj-pval
function/pathway	# of annotated genes in function	# of genes from associated loci annotated with function	# of connected graph(s) identified	Third column / Fourth column	statistic used in MEAGA	p-value	adjusted p-value for multiple testing

MEAGAtree provides the link between the genes (second and third columns) for the steiner tree(s) constructed for each function/pathway (first column)

Plot

If you want to visualize the shortest paths between the genes from the associated regions in a particular function/pathway, we provide a python function (it requires the matplotlib module from python) to generate plot (genes from associated regions are in red; other genes present in the shortest paths are in blue):

./bin/plotFunIntPPI.py -a ./db/gene2fun.txt -s ./db/splitFun_fungenesSP_BioGrid/R/REGULATION_OF_RESPONSE_TO_BIOTIC_STIMULUS -p REGULATION_OF_RESPONSE_TO_BIOTIC_STIMULUS -m ./example/marker2gene.txt -i ./example/intmarkers -o ./test/testMEAGA_plot

This will generate a figure for the function "regulation of response to biotic stimulus":

If you need help for this function, you could type:

./bin/plotFunIntPPI.py --help

Pre-calculated shortest path distance

The construction of Steiner trees is time consuming. The Kou’s algorithm requires the shortest distances and their paths to be first computed between genes. For effective performance, we pre-computed all shortest paths between genes in each function/pathway gene-set using data downloaded from different data sources and stored them in database to be readily retrieved when performing the Kou’s algorithm. We provide users copies of the compiled databases of shortest paths for interaction data obtained BioGrid (http://thebiogrid.org/), HPRD (http://www.hprd.org/), and STRING (http://string-db.org/).

If user wants MEAGA do perform enrichment test using the custom-interactome data (e.g. from other PPI source, or interaction data obtained from co-expression or text-mining analysis), we also provide a function to pre-compute all shortest paths for the custom data:

./bin/makeSP2Fun.py -i custom-interactome -g ./db/gene2fun.txt -o ./db/custom_splitFun_fungenesSP/ -f -m 3 -M 1000

custom-interactome is a two-column file indicating the gene pairs with biological association. gene2fun.txt is a file annotating the gene to function/pathway relationships (see above). The (-m and -M) parameters specify the minimum and maximum number of genes in the functions/pathways to be considered, respectively. By default (-F) computation of shortest path between genes within a function/pathway would use all genes in the interactome, if user only uses genes within the pathway, (-f) could be specified.

Tutorial

Using the markers from the Immunochip and the genetic association results from a meta-analysis (http://www.nature.com/ng/journal/v44/n12/abs/ng.2467.html), we provide the --marker2gene and the --intmarkers example files:

./example/marker2gene.txt
./exampleintmarkers

An example script for running MEAGA is also provided in ./example/testscript (run under the ./example directory)

../bin/MEAGA.py -s marker2gene.txt -g ../db/gene2fun.txt -i intmarkers -d ../db/splitFun_fungenesSP_BioGrid/ -o ../test/test -a 2 -m 5 -M 500 -t 10 -n 5000

The result files of the above scripts are stored in ../test/ with prefix "test".

It is not uncommon to observe genes coming from the same locus being annotated with the same function/pathway, if you want to restrict the analysis to functions/pathways overlapping with the associated genes which are all coming from different loci, you could use the provided function:

./bin/trimFun_NumIntFunGenesLoci.R ./example/interestedregionsgenes ./db/gene2fun.txt ./test/testMEAGAresult 0 ./test/testMEAGAresult_unique

"interestedregionsgenes" is an one-column file with each row represents one unique locus, and the associated genes within the locus are separated by the semi-colon. "0" represents the difference between the number of function/pathway- overlapping genes and the number of function/pathway- overlapping loci we want.

Citation

Lam C. Tsoi, James T. Elder, Gonçalo R. Abecasis. (2015) Graphical algorithm for integration of genetic and biological data: proof of principle using psoriasis as a model. Bioinformatics

Contact

If you have any questions, please contact [Alex Lam C Tsoi].