#C0493FQQZ3P Hi, Recently, my lab member asked me to perform

isoform

analysis. Previously, he generated

PacBio long-read RNA-Seq

data for lung adenocarcinoma samples and identified a number of

isoforms

. In the recent discussion he proposed to leverage our

Illumina short-read RNA-Seq

data (paired-end 150bp) to expand this analysis. This short-read data, also generated by him, is focused on examining similar conditions.” To explore further, I was referring this research article titled

rMATS-turbo: an efficient and flexible computational tool for alternative splicing analysis of large-scale RNA-seq data

https://www.nature.com/articles/s41596-023-00944-2 and found that computational tools for short-read RNA-seq analysis of alternative splicing can be classified into three major methodological categories:

isoform based

,

exon based

and

event based

.

Computational tools for short-read RNA-seq analysis of alternative splicing can be classified into three major methodological categories: isoform based, exon based and event based. Isoform-based tools (e.g., Cufflinks38, kallisto39) estimate the abundances and proportions of full-length transcript isoforms using short-read RNA-seq data, and can be used to detect novel transcript isoforms as well as isoform switches between biological conditions. However, transcript isoform discovery and quantification using short-read RNA-seq data are computationally challenging, as short-read RNA-seq only examines transcript fragments but cannot completely render full-length transcripts14. Moreover, many genes contain multiple alternatively spliced regions, and it is not trivial to associate the observed isoform switches to specific exons or splice sites. Exon-based tools (e.g., DEXseq40) use exon counts to detect differential exon usage, but do not provide information about the alternative splicing events that cause changes in exon counts. Due to these limitations, the third category of event-based tools represents the most widely used strategy. Many computational tools have been developed for event-based alternative splicing analysis using short-read RNA-seq data (for a nonexhaustive summary, see refs. 9,13,14). Some notable examples include classic tools developed on small-scale datasets in the early days of RNA-seq, such as MISO19 and SpliceTrap41, and contemporary tools that scale up to large-scale datasets, such as SUPPA/SUPPA242,43, MAJIQ44, LeafCutter45 and our own rMATS/rMATS-turbo18.

nf-core rnasplice

pipeline seems very comprehensive and leverages most of the tasks/methods mentioned above. I’d like to hear your thoughts what would be the best way to approach this using the nf-core rnasplice pipeline. What is right approach

isoform

based,

exon

based and

event

based?

Looks like you found some good literature to explain the differences. When you have reads of one or more kilobases, you can typically determine all or most of the transcript, which makes it easy. With short reads, that does not work as well, but the methods that try take the annotated transcripts and try to decide which ones are present in which proportions. There are also software packages that attempt infer transcripts (including unannotated ones). As you might imaging, all that has limitations with short reads. The methods that focus on single exons (e.g. DEXSeq) or splice junctions around exons/introns (event-based methods) are typically more sensitive since they operate on a scale where short reads are meaningful, but they tell you nothing about the rest of the transcript. What to use and how depends on your goals. If you have information about the transcripts from the PacBio experiment, you might be able to use those as the gene annotation that you use with the short reads, which might enable you to find out to what extent they are present there. Start with formulating the question you want to answer, break that down into smaller subquestions, then identify the methods that can help you answer each one.

@Thomas Danhorn Thanks much. This makes sense.