[ensembl-dev] SIFT score differences.

Will McLaren wm2 at ebi.ac.uk
Fri Jan 8 09:35:38 GMT 2016


Hello,

Differences in scores can be accounted for by two main factors:

1) the version of SIFT used to generate the score
2) the sequence set used to generate the alignments behind the scores

This also applies to PolyPhen scores.

It is not clear to me from the dbNSFP README (1) which version of SIFT they
use - they say "ensembl 66" as you noted, so perhaps they extracted the
SIFT scores from Ensembl v66.

For release 66, Ensembl ran SIFT version 4.0.5 using UniProtKB (release
2012_01, both the SwissProt and TrEMBL sets) (2); for release 75, we ran
SIFT version 5.0.2 with UniRef90 (release 2012_11) (3).

So it seems the scores presented by dbNSFP are generated by an older
version of SIFT, and with a different protein DB.

Hopefully this explains the differences for you

Regards

Will McLaren
Ensembl Variation

1: https://drive.google.com/file/d/0B60wROKy6OqccXV4UXNyNkJwNUk/view
2:
http://may2012.archive.ensembl.org/info/docs/variation/predicted_data.html#sift
3:
http://feb2014.archive.ensembl.org/info/genome/variation/predicted_data.html#sift

On 8 January 2016 at 05:59, binit treesa <binit.treesa at gmail.com> wrote:

> Hi team,
>
> Recently we found a source dbNSFP which is similar to VEP result.
>
> Usually we run VEP with variants to get the SIFT and polyphen score and
> its prediction.
> While comparing these two files, I noted that there is slight difference
> in the sift and polyphen score but the prediction remains same:
>
> The version we compared :
> dbNSFP  : it support ensembl v66
> VEP result  : VEP tool v75 (ensembl v75)
>
> It will be good if you can point out the reasons behind this score
> differences.
> Also please share the procedure to calculate the SIFT and Polyphen.
>
> Thanks,
> Treesa Binit
>
>
> _______________________________________________
> Dev mailing list    Dev at ensembl.org
> Posting guidelines and subscribe/unsubscribe info:
> http://lists.ensembl.org/mailman/listinfo/dev
> Ensembl Blog: http://www.ensembl.info/
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ensembl.org/pipermail/dev_ensembl.org/attachments/20160108/76014c4c/attachment.html>


More information about the Dev mailing list