[BioC] problems with strand in predictCoding

[Alex Gutteridge]

On 20.04.2012 02:59, Sean Davis wrote:
> On Thu, Apr 19, 2012 at 9:46 PM, Jeremiah Degenhardt
> <degenhardt.jeremiah at gene.com> wrote:
>> >From my perspective the proper behavior of predictCoding with 
>> respect
>> to strand would be to treat unstranded GRanges as positive strand as
>> this is the reference strand for things like the genome builds and 
>> vcf
>> files. Then the variants should overlap positive and negative strand
>> genes and should be reverse complemented for consequence prediction 
>> on
>> the negative strand genes. It should also allow overlap of negative
>> and positive stranded variants with both negative and positive
>> stranded genes, but properly reverse complement the variant in each
>> case to get the proper consequence.
>
> I agree with Jeremiah that the treatment of unstranded variants 
> should
> be to default to treating them as on the positive strand and reverse
> complement them for negative strand genes.  All other variant
> prediction softwares that I know of make this assumption and the VCF
> format itself seems to make something of an implicit assumption on
> this point.
>
> Sean

+1, also agree this would be helpful (and a reasonable assumption to 
make).

-- 
Alex Gutteridge