bio-basket (was RE: [BiO BB] how to access database remotely)

Dan Bolser dmb at mrc-dunn.cam.ac.uk
Thu Jan 1 13:28:42 EST 2004


++ HKG--
> Hi Dan,
>
>> I like the 'shopping basket' approach to bio data analysis.
>
> Would you like to elaborate a little bit on this design approach?


I am not sure how implementation details will pan out, but I imagine an 'object
basket' which we can carry between different pages and manipulate in various ways as
we go.

As the concrete details are a little shaky, here is an example use case to
illustrate what I have in mind.

1) Issue a query via the web interface for bacterial transporters (for example).

2) Dump the results in our basket as 'sequence objects'.

Now we want to know a bit more about the contents of the basket.

3) Navigate to a 'sequence clustering' page or a 'family database' page to discover
how many distinct families of bacterial transporter we have.

4) Add the family objects to the basket.

The family objects act over the sequence objects in a 'sensible way' as well as
carrying additional (family specific) information. For example each family may
encompass a much broader superset of sequences not annotated as bacterial
transporters.

5) Remove some stray families from the basket.

6) etc, etc.


I think by adding code to the basket as well as data we allow different web service
designers to integrate in a much more flexible way - but this is still shaky.

For example, someone could build a website to visualize a 'sequence alignment'
object, and carrying our basket to that page (after somehow acquiring alignment
objects (we dont care how)) we can see our data. Depending on the complexity of the
site we may be able to perform visual queries over our basket, before taking
analysis further.

As you can see this isn't an original idea, but I think the basket is a nice way for
people to feel like they have hold of something. Everything can take place over the
WWW and the basket itself can return a list of actions performed on its contents.
This 'action list' could then be printed, saved or exported to the web as a new
unary 'basket function' itself.

I love the idea that any analysis published in bioinformatics could be repeated by
any interested party in this way.

I feel this is a bit more 'open' than the SRS type system, and allows more
'sensible' integration more easily than other systems. However, we are talking
serious computer science technology under the hood.

The basket itself is where all the integration takes place, and so the objects it
contains (data objects or analysis objects) should have configuration interfaces to
allow this to be fine tuned by the user (not hidden from him).

Cheers,
Dan.









More information about the BBB mailing list