CCBB © 2018
Package gsia supplies functions described in the paper Wallace et al. 2018, On Entropy and Information in Gene Interaction Networks.
Dependencies: Matrix, igraph, mygene
First, install this package in your R by typing from command line,
sudo R CMD INSTALL gsia
Then, start R and type
library(gsia)
These functions are in gsia:
get_STRING
get_corr_matrix
get_Iij
inherit_interactions
tI
I
VI
IQR
get_p_tI
get_p_I
You can find out about them by typing ?get_STRING etc.
The examples are executable. For instance, in ?I you will find
df0<-as.data.frame(matrix(c(1,2,2,3,3,4,4,5,5,6,4,7,5,7,3,7,7,6,3,8,8,9,6,10,1,11,2,7,11,13,11,14,11,15,11,16,11,17,11,18,11,19,11,20,19,20,20,21,2,22,6,2,17,8,9,23,23,24,12,24,23,25,25,26,25,27,27,28,29,30),ncol=2,byrow=TRUE))
vnames<-as.character(1:30)
g<-graph_from_data_frame(df0,directed=FALSE,vertices=vnames)
set.seed(1357)
plot(g,vertex.size=11)
G<-get_corr_matrix(g)
A<-c("2","4","6","30")
B<-c("21","23","29")
I(A,B,G)
If you want to work with STRING network, you need to create the graph g_STRING and matrix G_STRING as follows:
g_STRING<-get_STRING() #this is for a human. You can get other organisms as well
If you have a tissue-specific gene expression profile, I recommend using the expressed genes to create a tissue-specific graph. Assuming expressed.genes is a character vector containing Entrez gene IDs of expressed genes, you can do that like this:
g<-inherit_interactions(g_STRING,expressed.genes) #some genes may become isolated vertices in this graph
You need to calculate the correlation matrix from your graph:
G<-get_corr_matrix(g)
If you skipped the tissue specific step, you can run this instead: G<-get_corr_matrix(g_STRING)
Now you can iterrogate actual gene sets. For instance, let's say that you want to look at a gene set (Alzheimer's Disease)
AD<-c("10347","10858","1139","1141","1191","1200","1378","1392","1471","1565","1636","1718","1808","1965","2","2023","2041","2099","2147","22976","23237","23385","23607","23621","2623","26330","267","27328","274","2932","3077","3162","323","3416","3479","348","3480","3481","3482","351","353128","3553","3630","3643","3952","406938","4129","4137","43","4353","450086","4524","4846","4852","498","51338","5328","54209","5468","55676","5621","5649","5663","5664","5697","581","590","596","627","642","64851","6517","6648","6653","7018","7019","7124","7167","7422","7447","7782","780912","801","8081","8301","836","945")
Could AD be a random gene set?
get_p_tI(AD,G) #about 0.003
You could add a second gene set (Inflammation):
IN<-c("10019","10544","10747","11221","11277","1137","1141","114131","114548","1241","1280","1395","1401","142","1437","177","183","1906","1947","1958","196","2056","207","2149","2150","2247","2588","2625","2833","284","2920","30009","3146","3162","335","3383","3458","3552","3553","3557","3569","3576","3586","3596","3605","3620","3665","3717","3934","3952","40","4057","406999","407040","41","412","4282","4313","4318","4353","4615","4803","4843","4879","4982","5020","51083","5244","5329","540","54106","5465","5468","54982","55829","5734","5743","6097","623","624","627","6347","6348","6351","6356","6584","6647","6774","6863","7018","7037","7039","7040","7076","7097","7099","7124","7349","7356","7422","7442","7538","796","834","885","8942","8989","916","9311","9370","9373","9536","9948","9966")
get_p_tI(IN,G) #about 0.000
Finally, you can ask if IN is related to AD:
get_p_I(AD,IN,G) #about 0