RNA-Seq 50 bp vs 100 bp

[Bardj]

Hello All,
I am getting into more and more RNA-Seq as our Core Facility is getting more requests, and a common question we have been encountering is if it is better to do RNA-Seq samples with 50 bp or 100 bp sequencing on our Illumina HiSeq2000.

From a Bioinformatics perspective, is there one clear advantage to doing one or the other? Do you gain valuable information from the extra 50 bp? Any thoughts on this would be greatly appreciated. Thanks

[peromhc]

One immediate thought: It dependes a lot on whether there is a reference available-- or if it is a de novo job..

[Bardj]

Say just for the sake of argument, that it was not a de-novo job. Say it was just a standard human build, like hg19.

I would likely not take on any de-novo jobs at this point, just due to the more complicated nature.

[NextGenSeq]

Alternative splicing is easier to detect with 100bp runs. Paired end 100 cycle runs makes it even more sensitive at detecting alternative splicing.

[dwmohr]

Not to complicate the issue, but I would think 50bp PE would be more desirable than 100bp single end.

EDIT: I should mention that CASAVA does not support PE RNAseq data...

[Bardj]

EDIT: I should mention that CASAVA does not support PE RNAseq data...

I was actually just looking this up as I saw the notification in my email, thanks for saving me time!

Do you guys have much experience with the CASAVA 1.8 RNA-Seq analysis pipeline? I've only used it a little bit. Is it something I can count on, or should I be running alternative workflows on the side?

[NextGenSeq]

Who uses CASAVA for anything? Once you have the FastQ files you can use a real program to analyze the data.

[Bardj]

Who uses CASAVA for anything? Once you have the FastQ files you can use a real program to analyze the data.

That is kind of how I've been operating so far, I typically utilize the conversion script to get into fastq files, and then run a basic eland alignment off in the background to get some statistics in the nice little charts.

After that I switch over to other tools.

[NextGenSeq]

Not to be snarky about Illumina's software but ELAND is not a very good alignment program either. Personally I prefer BWA.

[Bardj]

Not to be snarky about Illumina's software but ELAND is not a very good alignment program either. Personally I prefer BWA.

The only real thing I do like about it is the Statistics html file it generates with all of the alignment statistics, beyond that I haven't really utilized any of its output. We have been utilizing Bowtie for the most part, though I've been exploring other options such as BWA as well.

Do you have any recommendations for an RNA-Seq analysis workflow? Right now I've been using the tophat->cufflinks, but have been having memory issues with cufflinks lately with the larger sequencing files coming off the hiseq now with v3 chemistry.

[peromhc]

For me, the 1st step is always a vigorous cleaning of the reads.. I have been using The preprocess module of SGA, which implements the BWA model of trimming.. It improves denovo assembly quality, and reduces the number of unique kmer, and thus memory.

Especially with HiSeq v3, why waste resources on poor quality reads, when there are SO many high quality..