Best workflow for identification of isoforms and splice variants

Question

[Vojtěch]

Hi, is there an established pipeline or workflow for identifying and detecting isoforms/splice variants from bulk RNA-seq experiments? I am looking for some time without any success.

Answer 1

[JustinZhang]

For isoform quantification, please refer to these articles[1,2].

For alternative splicing detection: Sorry but none.

EXPLANATION:

1. The classification methods and detection results among splicing event callers are very different.
2. Also the results from next-generation sequencing and long-read sequencing are different, too.
3. There have been studies benchmarking pipelines based on limited samples or in-silico datasets, but the results are not good.

Till now, there is no such an universal & robust pipeline can give you reliable results as you want.

DETAILS:

1. Minghao Jiang et al. explained the major differences among AS callers and benchmarked many of them at event level, and an improved protocol was given. But, it was mainly based on simulated dataset. It's very different to detect events from clinical samples and in-silico fastq files. So be careful, its actual performance is not guaranteed.
2. Didrik Olofsson et al. discussed the performance of AS callers, and the results were not optimal. Even the Nextflow combing rmats and Whispering outperformed others, their absolute performance has a great room for improvement.
3. Amit Fenn et al. provided an integration tool called DICAST, comprising 11 splice-aware mapping and 8 event detection tools, but it's more like a framework rather than an ultimate solution. The authors pointed out that "we still find much room for improvement since tools with high recall values" and "the simulated data sets might not reach the same level of complexity as real biological data sets".
4. Xueyi Dong et al. showed that StringTie2 and bambu were the best choices for long-read sequencing data, but their precision and recall were all lower than 0.8.
5. Yaqi Su et al. got similar results with Xueyi Dong et al. And the IsoQuant was good.

OK, it would be too much for an answer on biostars. You can read articles yourself if you want. let's stop here and conclude them:

1. The tallest person you can find among dwarfs is, of course, still a dwarf.
2. The languages of dwarfs are not mutually accessible. More researches like DICAST are needed to build the Babel.

[dsull]

I think it would be helpful to the original poster if you linked to or cited some of those studies.

[JustinZhang]

Ladies and gentlemen, the references and necessary information have been added.

[dsull]

Great, thanks!

[Vojtěch]

I think I came across some of these, but I would still very much appreciated few of them cited. Thank you!