How to determine the proportion of overlapping genes in a genome?

[a.kmg]

Hello to all,

I'm working on stranded RNA-seq data and I would justify the use of this type of protocol, showing the proportion of genes that overlapp on the genome (human). Indeed, thanks to strand of information, it is possible to assign read to the real gene, for better quantification. Does anyone has a bioinformatics solution to determine this proportion ?

Thank you.

[cascoamarillo]

Hi,
Try bedtools intersected, it would do the trick finding the overlaping sequence.
Of course stranded RNA-seq is great! You don't need to justify it

[diego diaz]

If you are acquainted with R, maybe you should use the GenomicRanges package. Try converting the coordinates of the genes to a GRange object, then use the function "findOverlaps", finally you have to remove those overlaps where the same gene is compared.

[a.kmg]

Hi,

Thank you for your different ideas.

I used R to do that:

I imported my .gtf annotation file from ensembl:

# my ensembl gtf file
GTFfile = "Homo_sapiens.GRCh37.75.gtf"

#import libraries
library(GenomicRanges)
library(rtracklayer)

# importation of this gtf file in R
GTF=import.gff(GTFfile, format="gtf", asRangedData=F, feature.type="gene")

GTF_ok=as.data.frame(GTF)
dim(GTF_ok)
#57,773 initially genes

# separation of genes from plus strand and minus strand
GTF_strand_plus=GTF[which(GTF_ok$strand == "+"),]
GTF_strand_minus=GTF[which(GTF_ok$strand == "-"),]

##28,997 + 28,776 = 57,773 initially genes
dim(GTF_strand_plus)
#[1] 28,997 14
dim(GTF_strand_minus)
#[1] 28,776 14

# find overlaps between genes from plus and minus strands
overlap_with_strand=findOverlaps(GTF_strand_plus, GTF_strand_minus, ignore.strand = TRUE)

dim(overlap_with_strand)
#16,249 overlapps

My original GTF file contained 57,773 annotated genes.
After these commands, I found 16,249 overlaps, that to say, ~28% of overlapping in human genome (16249*100)/57773 = 28%. Or we consider one overlap as two genes and so : (16249*2*100)/57773 = 56 % of overlapping genes ?

This method is fine for you ?