generateEvents misses splicing events

Dear SUPPA developers,

We want to analyse data from Arabidopsis with your nice tool. The gene 'AT1G45474 have splicing events but they are not detected by SUPPA.

Do you have any idea why ?

Thanks a lot for your help.

Our call: python3 /bin/suppa/suppa.py generateEvents -i SUPPA.gtf -o generateEvents.gtf --event-type SE SS MX RI FL --exon-length 5 --boundary V --threshold 1

SUPPA.gtf ( only the rows for AT1G45474)

1	TAIR10	exon	17179302	17179537	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.2"
1	TAIR10	exon	17179610	17179742	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.2"
1	TAIR10	exon	17179828	17179940	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.2"
1	TAIR10	exon	17180021	17180217	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.2"
1	TAIR10	exon	17180297	17180439	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.2"
1	TAIR10	exon	17180701	17180806	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.2"
1	TAIR10	exon	17179302	17179537	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.1"
1	TAIR10	exon	17179610	17179742	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.1"
1	TAIR10	exon	17179828	17179940	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.1"
1	TAIR10	exon	17180021	17180217	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.1"
1	TAIR10	exon	17180297	17180530	0	+	.	gene_id "AT1G45474"; transcript_id "AT1G45474.1"

[EduEyras]

Dear Claes,

thanks a lot for your email. Looking at the exon coordinates of your gene, I can see that transcript .1 has the same exons as transcript .2 except for the 3’ end exon, which is a bit longer. Also, the 3’ end exons from .2 is missing in .1. Is that correct?

This type of variation is not considered in the definitions by SUPPA. As the variable exon in .1 is a terminal exon, its end coordinate cannot be used as an alternative splice-site. Regarding a possible alternative polyA, it is true that is a perfectly possible variation but SUPPA only considers the case when the alternative 3’ end exon is different, but share a common upstream 5’ splice-site.

Let me know if that is what you meant. Thanks for using SUPPA

E.

[EduEyras]

Hi,

On Thu, Oct 19, 2017 at 6:07 PM, Lilly88 [e-mail] wrote:

Dear Eduardo,

thank you for your reply. In my case I observe many more significant differentially used isoforms than events, just filtering for dPSI |0.1| and pval<=0.05 I find roughly 4 times more differentially used isoforms than events. Intuitively I would expect the opposite, but for what I understood the calculation of events is overly simplified, considering the usage of 1 exon in 1 isoform over its usage in the “closest” isoforms and not over all the isoforms with different first exons.

it can include all the others that conform to a definition of two alternatives. But in the case that alternative first exons it is true that it selects the isoform that gives the closest alternative exon.

There is a large number of alternative first exons (generic) and not all of them end up described as events in SUPPA. You probably get many more from isoforms for that reason. Also, they are generally very abundant genome-wide compared with other types of variations, at least in human.

For example, referring to the gene reported above the usage of the first exon of isoform 203 is calculated over the usage of exon 1 in isoform 201 and not over the sum of first exon of 201, 206+205, 204 and 202. As the denominator is smaller it is harder to reach significance. To me it looks like the definition of usage of first and last exon is made too simple to capture many splicing events which are presents. Is there a way to modify this?

The significance is tested at the level of differential splicing, and it could still be significant despite having a small denominator, unless you are using some expression cutoff.

We could try to include a variable number of exons between the two alternative exons to allow for these types of internal promoters. My worry is that for complex regions, this could end up producing an explosion of combinations that may obscure the interpretation of the result, whereas the isoform analysis would already be quite straightforward.

I hope this helps best

Eduardo

Thanks again!

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/comprna/SUPPA/issues/3#issuecomment-337956901, or mute the thread https://github.com/notifications/unsubscribe-auth/AMWVB35ikXPsMql3559oyPDtmRlxE1b_ks5st3O9gaJpZM4L6jeG .

– Dr E Eyras

ICREA Research Professor Universitat Pompeu Fabra PRBB, Dr Aiguader 88 Tel: [phone] (ext 1502) E08003 Barcelona, Spain Fax: [phone]

http://scholar.google.com/citations?user=LiojlGoAAAAJ http://www.researcherid.com/rid/L-1053-2014 http://regulatorygenomics.upf.edu/