Does MISO provide the RPKM/FPKM for each gene/transcriptome?

Question

[Yu]

Hi all,

I am a newbie to RNA seq analysis, some one told me that I can use MISO to get the transcriptome expression level. However, I done a example using MISO and found that there is nothing related to expression level in outputs.

Could anyone tell me if MISO provides the expression level of alternatively spliced genes in the outputs when it identifies differentially regulated isoforms/exons across samples? If MISO can do it, how to get the FPKM for each isoform by using MISO?

I know Cufflinks, eXpress, RSEM can quantify gene's expression level. Are there any other tools can do the same thing and provide a output of gene's expression level(RPKM/FPKM)?

Thanks

Answer 1

[Charles Warden]

MISO provides analysis of individual splicing events (exon skipping, intron retention, etc.), not gene / transcript expression levels (MATS would be another similar option).

The usefulness of tools like MISO and MATS is that I think they are more accurate than the whole transcript reconstruction. For example, all the RNA-Seq data that I have seen has very uneven coverage (for technical reasons), often with good coverage for only part of the read. Similarly, all examples where I have compared transcript-level expression and saw a potentially interesting result with one isoform behaving very different than the other isoforms didn't look reasonable when I visualized the alignment.

For gene expression analysis, I would use the gene-level quantification. cufflinks, RSEM, or eXpress should all be fine. They should all give you similar results (for gene-level expression):

http://cdwscience.blogspot.com/2014/02/mrna-quantification-via-express.html

[Yu]

Thanks to Charles. I need isoform-level expression quantification. Cufflinks, RSEM and eXpress would give me different results. I think there maybe big difference between the results.

[Charles Warden]

Overall, the concordance might not be that bad since I think most isoforms typically get similar results in differential expression analysis (however,I would argue that provides no benefit over the gene-level quantification).

For example, Figure 5B in this paper shows the overall concordance between cufflinks transcripts and RSEM-like transcripts (implemented in Partek).

The tricky part is that I would always check the alignments if you do get one isoform that is predicted to be very different than the other isoforms. In such cases, I've seen results that didn't look encouraging, but maybe you will get luckier.