Transcriptome vs. Genome alignments

[Mike2188]

I have some RNA-seq data back and I am performing alignments with Bowtie2. The organism I am working with has an available transcriptome assembly as well as genome assembly. I am not interested in splice junctions or discovering novel genes... should I align to the transcriptome rather than the genome?

[Brian Bushnell]

What kind of data do you have, and what is the goal of your experiment?

[Mike2188]

I have several laser microdissected subtissues of a specialized plant structure called the funiculus that connects to a developing seed. I am working with canola. I am looking for differential expression between these different subtissues to explain specific functions of this structure as well as the specialization of its subcompartments.

[Brian Bushnell]

I would absolutely recommend mapping to the genome, in that case, since tissues might differentiate based on differential splicing as well as raw gene expression.

You shouldn't use Bowtie2 for RNA-seq; an RNA aligner such as Tophat (the most well known), STAR, or BBMap will be able to handle reads mapping to splice junctions. If you do not handle reads that cross introns, your results will be wrong because they will be biased against short exons; and if you map to the transcriptome, your results will be biased in favor of the isoforms that are in the particular transcriptome you selected, which will NOT contain all isoforms. The results of transcriptome-mapping are also biased in favor of genes that have more isoforms, if all are in the fasta.

So, if you want data that is unbiased, considers differential expression of gene isoforms in different tissue types, and discovering novel genes, you should map against the genome. You stated that you are not interested in splice junctions or discovering novel genes, but I don't really understand where RNA-seq research that ignores such things would be useful, at present.

I was just at a meeting with PacBio today where RNA-seq data from a model organism (meaning, it had a well-established transcriptome, from Illumina sequencing) indicated the existence of roughly double the number of isoforms as were present in the existing transcriptome. Whether these hypothetical isoforms are real is unclear; but either way, that's real science. Mapping exclusively to a transcriptome is not, since the hypothesis dictates the results.