Alt Analyze: problem with calls of events

Question

[Lalla]

Dear all,

I am trying several tools to analyze alternative splicing. one of the most promising and reported to be very accurate is AltAnalyze. The interface is extremely user friendly and does not require coding skills. The output covers almost everything one could imagine: splicing events, associated changes in corresponding proteins, splicing of microRNA binding sites, etc. In my case it run overnight with little effort and I've got plenty of tables the morning after. Too good to be true. Exaclty. Indeed I quickly checked a couple of genes and I found some miscall of events. In one case the call was alternative N-term and retained intron, but the junction differentially used was between the first 2 exons and not between the 2nd exon and the intron...how can the program claim that there is a retained intron? In another case the splicing was between the first 2 exons and the corresponding protein would have and alternative N-ter but the program called alternative C-ter.

Did any of you have similar experience with this tool? Any suggestions on how to fix it?

For my analysis I used reads aligneo mm10 mouse genome with tophat2 and I simply downloaded the AltAnalyze reference Ens72 for mouse.

Thanks in advance for any help!

Answer 1

[Nathan Salomonis]

Hi Lalla,

If you haven't checked it out already, you can find extensive background information on the algorithms used for alternative splicing detection and interpretation in AltAnalyze here: http://altanalyze.readthedocs.io/en/latest/Algorithms/ http://altanalyze.readthedocs.io/en/latest/ProteinDirectionIndicator/

Publications using AltAnalyze can be found here: https://scholar.google.com/scholar?hl=en&as_sdt=0%2C36&q=AltAnalyze&btnG=

Also check out Figure 2 in the following paper that describes why you see domain predictions that wouldn't not be directly overlapping with a specific splicing event: http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1000553

Basically, when comparing two reciprocal junctions with the ASPIRE algorithm, the software is looking for pairs of minimally different RNA-isoforms (exon composition) that contain and don't contain the compared junctions. The associated proteins for those isoforms are then compared along with any protein isoform differences including retained introns. Some retained introns are actually very small (hundreds of nucleotides) and are effectively exons with exon IDs assigned by the AltAnalyze software rather than intron IDs. Such isoforms may include other exons that are different since some splicing events co-occur with others. This may not be the actual case, but rather a novel isoform is produced which only impacts that exons. For that reason protein isoform and domain interaction predictions are speculative. A tool that is distributed with AltAnalyze (AltAnalyze Results Viewer - Grey icon), lets you interactively explore such domain isoform graphs in the context of these spliced junctions and exons. http://altanalyze.readthedocs.io/en/latest/RunningAltAnalyze/#alternative-exon-visualization-altexonviewer http://altanalyze.readthedocs.io/en/latest/RunningAltAnalyze/#altanalyze-results-viewer

If concerned about the accuracy of these results, that is a good thing! Splicing results should be considered speculative and followed up with read-level visualization (Sashimi-Plot visualization in AltAnalyze or IGV). Importantly, the legacy splicing algorithms present in the current distributed compiled version of AltAnalyze are not ideal. We have been developing a new algorithm called MultiPath-PSI which is far superior in terms of sensitivity and specificity (see simulation validation and algorithm comparisons): http://altanalyze.readthedocs.io/en/latest/Algorithms/#multipath-psi-splicing-algorithm

You can apply this algorithm if use the latest version of the software from GitHub. The results will be stored as described in the above link. https://github.com/nsalomonis/altanalyze

A PyPi python installer is being finalized to install all dependencies on Mac and Windows here, but is still being tested:

pip install --extra-index-url https://testpypi.python.org/pypi AltAnalyze

(type "altanalyze" in a Mac terminal or Windows command prompt to start altanalyze)

Let us know if you need help with any of these analyses ([e-mail]).

Best, Nathan