Hi, everyone:

Thanks for your helpful replies. It is encouraging to know that the
problem already has a solution, I guess I just need to find out how
COSMIC is converted into a Biomart.

I understand that as long as data is transformed into 'reverse-star'
schema, it can be turned into a Biomart. But I'm not sure whether
reverse-star is suitable for all types of data sets. For each Biomart
dataset, my understanding is that it centers on a hub, an entity that
other tables are linked to through foreign-key relationships. For
COSMIC, I cannot find a single hub that links all the tables together. I
had to create multiple data sets, each centered around concepts such as
'sample', 'gene' and 'paper'.

It seems to me that COSMIC is more suitable for a 'star' schema with
'mutation' being the entity that holds everything together. If a data
set is more suitable for a 'star' schema, does this mean that it is not
suitable to be transformed into a Biomart?

Zhengyan (George) Kan

-----Original Message-----
From: Syed Haider [mailto:[email protected]] 
Sent: Friday, May 14, 2010 7:39 PM
To: Kan, Zhengyan
Cc: [email protected]
Subject: Re: [mart-dev] dataset not suitable for biomart?

Hi George,

BioMart reverse-star schema is independent of any biological concepts, 
hence can be used to model any kind of 'omics data. As far as the COSMIC

goes, could you please let us know what difficulties did you face ? we 
are happy to help.

Best
Syed

On 15/05/2010 01:47, Kan, Zhengyan wrote:
> Hi, Bio-mart team:
>
> I wonder if there are some data sets not suitable for transformation
> into Biomart. In my experience, I had difficulties transforming the
> COSMIC database into Biomart as the data is very decentralized. It may
> just be my lack of knowledge about the Biomart system though.
>
> My understanding is that Biomart uses a 'top-down', reverse-star
schema.
> Should this type of schema applicable to all genomic data sets or just
a
> limited spectrum?
>
> Many thanks!
>
> Zhengyan (George) Kan
> Computational Biology
> Pfizer, La Jolla
>
>

Reply via email to