Bowtie and ERANGE

[Bio.X2Y]

Hi,

We're planning on using the Mortazavi ERANGE package for quantifying gene expression in our rna-seq data. An aligner must be used for this process, and we are going to use Bowtie.

The ERANGE documentation describes how Bowtie should be used to create output that is compatible with the ERANGE process, but part of it is confusing me:

We use the following settings to emulate Eland output:

$BOWTIEDIR/bowtie zzz -v 2 -k 11 -m 10 -t --best -f s1.32mer.query.txt --unfa s1.unmapped.fa --maxfa s1.repeat.fa s1.zzz.bowtie.txt

where zzz is the genome prefix that you gave when building the
genome. In particular, we ask bowtie to map all multireads up
to 11 ("-k") with up to 2 mismatches ("-v" and "--best"), however
we will only import all multireads up to 10x multiplicity ("-m").

My understanding is that the "-m 10" will make the "-k 11" non-sensical, since any reads with more than 10 matches will be discarded, therefore making it impossible to report up to 11.

Have I misunderstood?

Thanks!

[Ben Langmead]

Specifying "-m 10" without "-k 11" would cause (a) all alignments for reads with more than 10 alignments to be suppressed, and (b) one random alignment to be reported for reads with 1 to 10 alignments. Specifying both (equivalent to "-m 10 -a") causes (a) all alignments for reads with more than 10 alignments to be suppressed, and (b) *all* alignments to be reported for reads with 1 to 10 alignments

Hope that helps,
Ben

[Bio.X2Y]

Hi Ben,

Thanks for the explanation.

Does this mean that "-k 10" or "-a" would have been more appropriate here, since "-k 11" would be as redundant as "-k 1234" ?

I suppose I'm trying to determine if I'm missing something from the ERANGE documentation. Does their explanation in blue make sense?

Thanks!

[Thomas Doktor]

I think you should also try out TopHat and Cufflinks as it does much the same thing as Erange.

In addition, TopHat and Cufflinks were (at least for me) much easier to set up, and the latest versions seem to have sorted out the most significant bugs. Never having succesfully run Erange I can't tell you if there are speed differences, but I think cufflinks is probably much faster than Erange (I believe much of the code is in Python, but there might be C/C++ code for the speed sensitive parts).

Looking at the list of Cufflinks contributors I see Ali Mortazavi and Barbara Wold, so Cufflinks might even have design elements in common with Erange. Ben, you must know a lot more about this?

[Ben Langmead]

@Bio.X2Y:

Yes, their explanation makes sense, (incidentally, it's similar to what TopHat does). I would have said "-a -m 10", but, as I say, the effect is exactly the same as "-k 11 -m 10" or "-k 1234 -m 10".

@Thomas Doktor

Yes, TopHat and Cufflinks are highly recommended. Ali and Barbara and Cole (and others) collaborate very tightly on them. Please check them out! If you have questions about those tools, feel free to post them here. Cole reads this forum with some regularity.

Thanks,
Ben

[Sparx]

Hi,
Could anyone explain to me where these input files come from
s1.32mer.query.txt
s1.unmapped.fa
s1.repeat.fa

I only have the fastq files downloaded from GEO-database and the bowtie d_melanogaster index files.
Thanks