If you happened to align against a reference genome from UCSC, then you can just download the refseq file from the table browser. This is the simplest method to get a BED12 file for computeMatrix.

If you aren't using a UCSC reference genome (we don't), you can construct the BED12 files as such (this is actually part of my reference genome Makefile):

1. Download a GTF file from Ensembl/Gencode/UCSC/etc.
2. I then use the the following, assuming the GTF file is called "annotation.gtf" and the BED file should be called "annotation.bed":

awk '{if ($$3 != "gene") print $$0;}' annotation.gtf | grep -v "^#" | /package/UCSCtools/gtfToGenePred /dev/stdin /dev/stdout | /package/UCSCtools/genePredToBed stdin annotation.bed

Practically speaking, I can also provide appropriate files for most common genomes (mouse, human, fly and worm). Since we developed deepTools, we have a lot of these and I probably just post whatever you need if you can't directly get it from UCSC.

you can get that particular output directly from Ensembl at http://www.ensembl.org/biomart/martview/

> select Ensembl genes, homo sapiens (reference dependent)
> go to the attributes section
> unselect ensembl gene id, transcript id
> select chromosome, gene start, gene end, associated transcript name, associated gene name, strand
> export results

note that chromosome names won't have the "chr" suffix. add it if needed.