Best RNA-seq program for unpaired reads

[reshetovdenis]

Hi everyone,
I have RNA-seq unpaired reads from Solid machine. The read length is 35.
Please, help me: I can't find any software except ABI (that works on PBS and LSF clusters) to analyse the data. I do have access to a cluster machine but it has no PBS or LSF.

[kopi-o]

There are many aligners that can deal with SOLiD's color space format nowadays. Google for BFAST, BWA, Bowtie or Mosaik, for example.

[reshetovdenis]

kopi-o, thanks for reply. I've seen these programs and used Bowtie. But they are good for sequencing genomes not for RNA-seq data, am I wrong?

[Richard Finney]

If the data is small enough, use blat. Access to a biowulf cluster can help if it's big.
This will take a long time.
You can prefilter using a SR aligner (e.g.: bwa).
Map to genome. Map against refseq. Merge outputs. (thest steps are hard to do).
Then blat the ones that don't map ... and merge again.
Not easy, but doable.

[mrawlins]

What we've used on our SOLiD data (50bp reads) is the BioScope WT analysis pipeline. I am impressed with BowTie from a technical perspective, and it seems to give good alignment, but it isn't able to do junction mapping (i.e., you'll get no splicing information, and all reads that cover a splice junction will go unmapped, or be mapped erroneously).
I haven't seen any software able to do the junction mapping in colorspace other than BioScope or the Corona Lite WT pipeline, from ABI. Some people are comfortable using the "double-encoding" trick and using any old junction mapper, but I remain unconvinced of the accuracy and reliability of such a method. If I didn't need splicing information, I would go with BowTie. Since I sometimes do, I'm sticking with ABI's software.

In my estimation it's worth getting TORQUE/maui up and running on your cluster and using ABI's software.

And if you find something else that will do junction mapping on SOLiD reads, I'd love to hear about it.

[kopi-o]

As far as I know, though. Bioscope uses a splice junction database to align against, rather than finding splice junctions de novo, so it is not really conceptually different from using another aligner and adding artificially constructed exon-exon junctions to your reference genome (although that takes some extra effort, of course). The previous versions of the ABI pipeline, though, chopped up the reads into two pieces and mapped them separately, which enabled de novo splice junction detection.

[jwfoley]

kopi-o, thanks for reply. I've seen these programs and used Bowtie. But they are good for sequencing genomes not for RNA-seq data, am I wrong?

Other way around. These are short-read aligners that simply map each tag against the reference sequence; they're designed for quantitative, targeted genomics like RNA-seq or ChIP-seq, but basically useless for de novo assembly.

Bowtie and BFAST can both handle SOLiD data. BWA's treatment of colorspace is an ugly hack and I wouldn't use it.

[Pejman]

And if you find something else that will do junction mapping on SOLiD reads, I'd love to hear about it.

TopHat-ers say that TopHat's is gonna be able to handle colorspace data very soon.

[zee]

Try NovoalignCS (www.novocraft.com) it's quite sensitive and has many useful features for SOLiD reads. There's also documentation on how to do alignment to a full genome.
NovoalignCS will do mismatches and indels with colorspace reads and supports csfasta/qual formats.

[Pejman]

Seems nice but I guess NovoalignCS needs a license.

[zee]

Yes it does but it's free and IMO worth a try if you're evaluating accurate tools for your work.

[Cole Trapnell]

TopHat-ers say that TopHat's is gonna be able to handle colorspace data very soon.

We released a version of TopHat that supports Colorspace yesterday (10/3)

[Pejman]

We released a version of TopHat that supports Colorspace yesterday (10/3)

wow, cool I knew that it's gonna be released soon but, cool! We are not interested in new splice junctions, but Bowtie output does not fit to Cufflinks, so why not TopHat.
I'm now trying out NovoalignCS, do you have any comparisons with other pipelines already? Let's say assuming that we don't care about novel splice junctions, just simple RNAseq for differential expression analysis.

[jgibbons1]

I really like the program rSeq by Hui Jiang.

1st convert your reads to standard fasta format. The software has a mapping program and then a script that calculates RPKM for each reference transcript.

Also...its free!

[danielr]

wow, cool I knew that it's gonna be released soon but, cool! We are not interested in new splice junctions, but Bowtie output does not fit to Cufflinks, so why not TopHat.

You get the right output format with the -S option in Bowtie (TopHat has nice splicing-related features Bowtie lacks though).