Raw counts of 12 column bed file (against multiple BAM)

[swaraj]

Hello Everyone,

I am sharing a perl script to get raw read counts from mulitple BAM files, for a given bed file. The script requires bedtools and samtools to be preinstalled in the system.

I hope it is useful for users. An alternative solution exists in python
(http://seqanswers.com/forums/showthread.php?t=18182)

[jason_ARGONAUTE]

thanks so much for ur script.
But you didn't give the usage info.

I hv a .gtf file download from EMBL for D.mel now, should i convert it to .bed in advance?
and the usage is?
perl bed_count.pl annotated.bed dir_of_my_bamfiles tmpcount/ my.exon my.count

[swaraj]

Dear Jason,

A much quicker solution is to

1. Convert GTF to BED. A nice script is found here http://ea-utils.googlecode.com/svn/t...lipper/gtf2bed

2. Convert the BED file to exon coordinates (BEDTools package).
bed12toBed6 -i filename.bed > filename.exons

3. Get read counts from multiple BAM files (BEDTools package)
multiBamCov -bams aln.1.bam aln.2.bam -bed filename.exons > filename.counts
(Remember to build index of each BAM file using samtools index, else multiBamCov wont work)

4. Load the count file in R environment.

filename <- read.delim("filename.counts", header=FALSE,sep = "\t", stringsAsFactors =FALSE)

5. Sum the exons into gene counts using apply function in R
input <- filename[,7:ncol(filename)]
names <- filename$V4

filename.gene.count<-apply(input,2,function(x){tapply(x,names,function(y){sum(y)})})

6. Write the count file as a text file
write.table(filename.gene.count, file = "filename.gene.count", quote = FALSE, row.names=FALSE, col.names=FALSE)

The order of the counts for samples will be the same order in which you give the BAM files. My previous script is easier to implement but little slower. If you open the script in a text editor you can give the name of the directory of BAM files and the BED file. Hope this helps.

[jason_ARGONAUTE]

Dear swaraj,

It helps me a lot. Thanks again^^
But i hv encountered a now problem now when adopting the new solution. I hv got 23017 exons in total while 14797 genes are mapped to the genome by Tophat & Cufflinks. and 14797 is a proper count for fruit fly. I notice that ur solution use exon coordinates(FBtrxxxx instead of FBgnxxxxx).
I am confused now. what else do i need to do to process the matrix of raw counts i just got before importing it into R packages like edgeR for DEG analysis?

[swaraj]

In step 2 I coverted transcript BED file into exon file. Here the name of each exon is the transcript name which means transcript name is reduddant. In step 5 this same file is converted back into transcript file using R. All reads in exons with same transcript name are summed up. I would not go into using exon names. Just keep the same name (gene name or trasncript name) for each exon of a feature. This makes it quicker in R to sum the counts of all exons.

[jason_ARGONAUTE]

ya, i know and i do exactly as ur pipeline. might be i mixed the concept of exon and transcript. i thought they were the same thing.(both named "FBtrxxx")
So, this is not the reason why i got almost twice counts of genes as i expected.
Could you figure out any other reason? Thank you!

[jason_ARGONAUTE]

Or,
what do i need to do to get raw counts for each gene?
Thank you!

[dpryan]

Or,
what do i need to do to get raw counts for each gene?
Thank you!

If you have a GTF file, just use htseq-count.

[jason_ARGONAUTE]

I hv tried htseq-count, really hard.
But it cannt be installed on our lab server.
Thank you all the same^^