[ensembl-dev] Prediction of consequence type for novel variants

[Will McLaren]

Hi Sung,

There is a publication referring to the system, but it does not go
into great detail on the internal workings:

http://bioinformatics.oxfordjournals.org/content/26/16/2069.abstract

Here's an approximate flow of what happens in the API. The vast
majority of the code used is in the Core module
Bio::EnsEMBL::Utils::TranscriptAlleles.pm, mainly the methods
type_variation() and apply_aa_change():

- find overlapping transcripts (using $vf->feature_Slice and
$slice->get_all_Transcripts), then for each transcript:

- get transcript mapper and map variation's coordinates to cDNA, CDS and peptide

- any variants that don't fall in the coding sequence are classified
here (e.g. INTRONIC, UPSTREAM) and the flow ends

- if variation falls in exon (i.e. has defined CDS coordinates),
generate alternative codon(s) and resulting translation

- compare translation to reference; classify as e.g.
SYNONYMOUS_CODING, NON_SYNONYMOUS_CODING

We are currently working on an overhaul to this system which should
make it easier to comprehend by following the code.

I would recommend trying to follow through the code in Perl's
debugger, using the "perl -d" option.

Hope this helps

Will McLaren
Ensembl Variation

On 9 December 2010 13:19, Sung Gong <sung at bio.cc> wrote:
> Hi,
>
> I was thrilled to find that Ensembl API provides a nice script
> (ftp://ftp.ensembl.org/pub/misc-scripts/) which can predict the
> consequence types of novel variations.
> Also, good to see a good demonstration how to use the API for that purpose:
> http://www.ensembl.org/info/docs/api/variation/variation_tutorial.html
>
> Before realising the variation API can help predicting consequence
> type of novel variants, I used to use only core API to map the
> position of my variants to see whether they are within coding region,
> intron, exon and so on.
> Now, I wondered how the variation API works for that purpose - looked
> at the source code, but found it is somewhat overwhelming.
>
> Can anybody explain how the novel prediction works internally under the hood?
>
> Cheers,
> Sung
>
> _______________________________________________
> Dev mailing list
> Dev at ensembl.org
> http://lists.ensembl.org/mailman/listinfo/dev
>