[Gusdev-gusdev] Re: Alternative-splicing discussion

[Arnaud K. <ax...@sa...>]

Hi everyone

> Genes and alternative-splicing
>     Topic addressed was defining when transcripts are alternative forms
> of the same gene and when they are different genes. The most important
> point is that GUS is completely flexible in which RNAs are assigned to a
> Gene, Several cases were discussed as to how the different groups would
> make assignments. There was consensus in that RNAs should have some
> overlap of translated regions in order to be considered as deriving from
> the same gene. Overlaps on opposite strands or in intronic regions were
> not considered as grounds for being the same gene. For CBIL, it did not
> matter whether the proteins derived from alternative splice forms had
> different functions. For PSU, if proteins had minimal overlap and had
> different functions they were usually annotated as different genes. Both
> approaches are allowed by the schema. The schema does however require
> that an exon have one parent. Thus, if an exon is shared between two
> RNAs that are assigned to different genes, the exon will be represented
> twice, once for each gene. Also, if an exon is coding for one RNA and
> non-coding for another RNA, it will be duplicated even if the RNAs are
> from the same gene in order to specify the difference in coding.  This
> is because ExonFeatures contains a number of attributes specifying
> coding start and stop and reading frame as well as initial exon. For
> future, these attributes may be moved to another table such as
> RNAFeatureExon.
>
>  

I'd like to illustrate the alternative splicing discussion by the worst 
case scenario Bart could up with: the major immediate early (MIE) region 
of the  Human Cytomegalovirus (HCMV).

See attached the map of this region. Briefly, the upstream gene has 5 
exons and gives, after splicing, 5 transcript variants and consequently 
5 proteins whose size differs quite a lot. In this situation all exons 
can have both behaviour, coding/non-coding, depending on which RNA 
variant they're attached to.
Besides, note that the 5th exon can be splited into 2 smaller exons and 
an intron, so it's not only the behaviour of the exon which changes but 
it's internal structure as well !!!!

What we would like is the flexibility of either attaching the RNA 
variants to separate gene feature entries or to a unique gene feature 
entry. As long as we have this flexibility, the underlying schema is 
fine with us.
At the ExonFeature level, you're mentioning that the annotation of the 
exons (first/last/middle - coding/non-coding) could move to the 
RNAFeatureExon table. Would it replace the duplication of the exons ?

cheers
Arnaud

-- 
Arnaud Kerhornou

The Wellcome Trust Sanger Institute
The Pathogen Sequencing Unit
Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, UK
Work: +44 (0) 1223 494955
Fax:  +44 (0) 1223 494919