Transcript See Spot Run Building Web-Based Systems for Visualizing

VERTEX PHARMACEUTICALS INCORPORATED
See Spot Run:
Building Web-Based Systems for
Visualizing Large Data Set with Spotfire
Dr. Kenneth Tupper
Lead Scientific Software Analyst
Research Informatics, Information Systems
Spotfire Users Group Meeting
May 3rd, 2001
© 2001, Vertex, Inc. - All Rights Reserved.
www.vrtx.com
The Problem: “I’m seeing spots!”




The screening group generates
about 3 Gb of data per year.
Manual review of screening results
took several hours each day.
Retrieving data from ActivityBase
and reformatting the data for use in
Spotfire took several steps
No feedback between Spotfire and
ActivityBase to update database
How much is really ‘too much data’ ?



In 1992, the existence of a
black hole at the center of
M87 was confirmed using
the HST/FOS.
This HST study generated
2 Gb of data per day.
Data could not be stored
and was processed in real
time.
Data  Information  Knowledge




Graph of inhibition data for 8
proteins.
Too much data in this view to
extract useful information on
cross inhibition.
Data reduction techniques
(clustering, trellising) can be
used to extract useful data.
Data reduction should not be
done during initial data
retrieval
Who Will Win the Race?

Given a query where A returns 5,000 rows and B returns
50,000 rows, which query will run faster?
SELECT A, B FROM TABLE
SELECT A, B FROM TABLE
WHERE A = ‘Prot1’
AND B = ‘Valid’
WHERE B = ‘Valid’
AND A = ‘Prot1’
1.45 seconds

1.00 seconds!
Oracle resolves predicate list from the bottom up.
To Web, or not to Web...




Ease of development and use - simple web interfaces can
be developed for each functionality.
System can be deployed across different computing
platforms, assuming that no platform or browser specific
plug-ins or controls are used.
Lower maintenance costs - updates are performed on a
server, instead of on each desktop machine.
“If it looks like a Visual Basic application when you’re
done, you should have written it in Visual Basic” - KT
“Will it run on my Palm Pilot?”
Software Design Considerations





Users will not know the internal table structure of
ActivityBase or how to form complex queries using SQL.
The interface should be able to run on users’ desktop
machines (Macintosh at Vertex).
The system must be extensible to allow addition of new
features.
Database security must not be compromised for ease of
implementation.
KISSME: Keep It Simple, Secure, Maintainable, and
Extensible.
Oracle Application Server (OAS)


OAS uses a three-tier computing model to provide
database access to web clients.
Web applications can be developed using the following
cartridges:
–
–
–
–
–

PL/SQL (stored procedures)
Live HTML
Java (JWeb, JServlet)
Perl
C, C++
Security is handled by the web server.
OAS PL/SQL
A form defined as
<FORM action="http://pooh.vpharm.com/plsql/atest/passwd" method="POST">
<P>New Password:
<INPUT type="PASSWORD" name="npassword" size="40”>
<P>Confirm Password: <INPUT type="PASSWORD" name="cpassword" size="40”>
<INPUT type="SUBMIT" value="Submit">
</FORM>
can be submitted directly to a stored procedure
CREATE or REPLACE PROCEDURE passwd (
npassword in varchar2 DEFAULT NULL,
cpassword in varchar2 DEFAULT NULL)
AS...
Welcome to “Spot”
Acknowledgements





Mark Namchik (HTS Group)
Sergio Rotstein (Research Informatics)
Steve Schmidt (Information Systems)
David Nick (Spotfire)
Nick Gourju (ActivityBase)