Event-based software for alternative splicing with three conditions?

Question

[htyu1lmzta]

Hi all,

I am looking for a piece of software that can be used to investigate alternative splicing and fulfills the following four criteria:

• can handle more than two experimental conditions
• breaks AS events down by type (e.g., cassette exon, mutually exclusive exons, alternative 5’ splice site, alternative 3' splice site, intron retention)
• can be used in rat
• AS events can be mapped back to protein sequences

I am having a surprising amount of difficulty identifying a tool that meets these criteria. Exon-based tools like DEXSeq, diffSplice, or JunctionSeq can be used to analyze more than two groups, but cannot readily distinguish between different types of AS events. Conversely, among the event-based tools I have investigated, two (MISO and VAST-TOOLS) do not seem to be available for rat, one (rMATS) seems to only handle two-group comparisons, and one (MAJIQ) returns very complex events that I am having quite a lot of difficulty mapping to protein sequences.

Is there a method I am not aware of that could be used to identify alternative splicing events in rat in an experiment with three conditions?

Thanks in advance for your help.

[Jeremy Leipzig]

i'm not so sure software that handles 3 conditions is really much better than A vs B and A vs C and B vs C

[htyu1lmzta]

Why do you say that? It seems fundamentally different to do what amounts to an ANOVA as opposed to a series of two-condition tests.

[Jeremy Leipzig]

that's an interesting point. ANOVAs became popular before computers, and they require a two-way post-hoc anyway. The majority of efforts in differential expression are two-way, so it's no surprise differential splicing would be the same.

Granted you are introducing some multiple testing issues when you perform more than one two-way comparison. Then again you are already doing multiple testing when you are comparing more than one gene, so you will be doing some kind of correction in any case.

[Kristoffer Vitting-Seerup]

ANOVA-type of tests (F-tests) are also really useful for cases where you don't care about which of the pariwise comparisons it occurred in (e.g. timecourse)

Answer 1

[Nathan Salomonis]

The software AltAnalyze meets these different objectives, depending on the type of multi-group test you want to apply. Firstly, it can perform splicing analyses (and gene expression) directly from FASTQ or aligned BAM files (make sure you are using the correct AltAnalyze genome database version). BAMs will be better since you will have novel junctions and intron-retention. Alternative splicing calculated via a PSI-based method (MultiPath-PSI - https://altanalyze.readthedocs.io/en/latest/Algorithms/#multipath-psi-splicing-algorithm). Protein isoform and domain-level impacts are predicted along with splice-type (e.g., alt 3', alt promoter, intron retention). The software is easy-to-use and can be run from a precompiled GUI or source-code (Python 2.7). Currently, it does not compute ANOVA's for alternative splicing directly, but output files can be reformatted for ANOVA analysis in AltAnalyze. For assistance contact: [e-mail].

Answer 2

[Kristoffer Vitting-Seerup]

I don't think such a tool exist. Many tools exist which can do 3 of the 4 tasks (see overview of approaches here) but the problem is that doing functional analysis (mapping back to protein sequence) requires you know the entire open reading frame overlapping the splice site of interest - and that you per definition cannot (for most cases) unless you are doing the analysis at transcript/isoform level.

From transcript level analysis you can get easily predict functional consequences (domain loss/gain) and get splicing information (by comparing the two isoforms in a isoform switch) with tools such as my R package IsoformSwitchAnalyzeR (which provides analysis both for individual genes and at a genome wide level consequence and splicing level).

But I'm not aware of any tool which allows an ANOVA-type test (F-test) at transcript level since you loose the direction necessary for defining the switch in the process).

Please also note that most isoform switches contain multiple splice changes (multiple splicing sites and/or changes in transcription start or termination) - see e.g. figure 4 in Vitting-Seerup 2017.