Casava1.8.2

[cnicolet]

Hi:
How are people feeling about CASAVA 1.8.2? We are just starting to use it and so far are underwhelmed. The new formats are cumbersome and the way it delivers multiple files is complicating everything we do downstream. We also benchmarked the new ELAND aligner and it performed exactly the same as the old aligner (which in our hands only delivers about 80% of the mapped reads as bwa), despite claims of superiority. Maybe we're not using it correctly? Any tips?

[simonandrews]

There's supposed to be an option in 1.8.2 to generate only a single fastq.gz file which might make things a bit easier.

I'm afraid we gave up trying to work with Eland since 1.8 due to the extra complexity of the run commands. We were perfectly happy with the performance of the old Eland, but have now moved to bowtie for everything (which is also a pain as you can't run paired end gzipped data through it without uncompressing first).

[cnicolet]

Thanks Simon! We did find some info on that using the --fastq-cluster-count parameter. But it's a little confusing, the manual says "Specify 0 to ensure creation of a single FASTQ file" then later it says "If you need to generate one unique fastq gzipped file for use in a third-party tool, you can set the --fastq-cluster-count option to -1" But I agree with you about Eland, I think our days of using that are over too.

[BAMseek]

... but have now moved to bowtie for everything (which is also a pain as you can't run paired end gzipped data through it without uncompressing first).

Hi Simon,

If the pain is about unzipping the files into temporary storage first, I think there might be some ways to uncompress and feed into Bowtie on the fly.

One way that I think works would be with named pipes like this:

rm -f pipe1
rm -f pipe2
mkfifo pipe1
mkfifo pipe2
gunzip -c reads/e_coli_1000_1.fq.gz > pipe1 & \
gunzip -c reads/e_coli_1000_2.fq.gz > pipe2 & \
./bowtie e_coli -1 pipe1 -2 pipe2

You could also try uncompressing and merging the paired files on-the-fly into the special tab-delimited format accepted by Bowtie, where the columns are

read_name[tab]sequence_1[tab]quality_1[tab]sequence_2[tab]quality_2

and pipe that into Bowtie.

best,
Justin

[simonandrews]

Thanks for the suggestions Justin. I hadn't realised there was a paired end format already - I might just make up a wrapper for that.

I'm sorely tempted to just patch in gzip support to bowtie, but don't want to have the hassle of maintaining a fork in perpetuity.

[earonesty]

Bowtie 0.12.7 patch for gzip support:

http://code.google.com/p/ea-utils/source/browse/trunk/clipper/bowtie-gzip.patch