Gostats hypergeometric test for kegg enrichment pathway non model organism

Question

[claire.morandin]

Hello

I would like to run an hypergeometric testing on KEGG terms for non-model organism. When it comes to Goterm I always use Gostats where I can enter my "universe" list and then my "target" genes and it works really well.

However, when it comes to Kegg pathways, I tried to use go stats again but it's based on the kegg.db which is outdated and as a result I get the following error:

Loading required package: KEGG.db
Failed with error:  'package 'AnnotationDbi' 1.24.0 is loaded, but >= 1.25.15 is required by 'KEGG.db''
In addition: Warning message:
package 'KEGG.db' was built under R version 3.1.0 
Error in .testKEGGFrame(x, organism) : 
 None of elements in the 1st column of your data.frame object are legitimate KEGG IDs.

Does anyone have any recommendations for Kegg enrichment analysis for non-model organism and /or how to do an hypergeometric test with R directly?

Thanks a lot for your help

CLaire

Answer 1

[lkmklsmn]

An intuitive approach to conducting enrichment analysis is done by using a two-by-two table of counts:

              YESanno    NOanno
YESgeneset    X1         X2
NOgeneset     Y1         Y2

where X1 is the count of genes in your geneset AND in the annotation class (e.g. Goterm). X2 is the count of genes in your geneset AND NOT in the annotation class. Y1 is the count of genes NOT in your geneset (e.g. universe-geneset= background) AND in the annotation class. Y2 is the count of genes NOT in your geneset AND NOT in your annotation class.

You can give this table to the R function fisher.test().

e.g. fisher.test(matrix(c(X1,X2,Y1,Y2),nrow=2,ncol=2,byrow=T))

This function will give you the hypergeometric test for a given gene-set and given annotation class.

[claire.morandin]

I agree with you, however if I want to report only the enriched pathways, with this method how do I know which ones are enriched?

[lkmklsmn]

You can come up with you own criteria for enrichment. E.g. p-value < 0.05, odds ratio >1 etc...

Answer 2

[Arjun]

Hi Claire,

You can use iPathwayGuide for Go analysis as we use KEGG database. iPathwayGuide is a pathway analysis application based on Impact Analysis approach that considers the role, function, interactions between the genes on each pathway. As of now, you can work with human data in cuffdiff, DEseq and custom 3-column tab delimited file but in couple of weeks, you will also be able to work with mouse and rat data. We are rapidly adding features and functionalities to our application and soon the choice of organism will also increase.

Currently the iPathwayGuide application is free to use and you can register here. Please feel free to reach out to me if I could be of further assistance.

Best,
Arjun Gupta
arjun (at) advaitacorporation (dot) com
Linkedin
www.AdvaitaBio.com