<%
pageTitle = "Smokin' Supercomputing"
articleid = "4419"
localnav = ""
section = "Archive"
author = ""
contributer = "BIO-IT World"
publicationDate = "2004-02-18"
description = "Paracel's Jason Molle on solving the IT challenges of biological analysis "
idgurl = "http://www.bio-itworld.com/archive/021804/horizons_smokin.html"
copyright = "2004"
keywords = ""
adsite = "idg.us.bioitw.archives"
ad468x60 = "yes"
ad150x800 = "yes"
ad336x280 = "no"
adSpecial1 = "no"
adSpecial2 = "no"
adSpecial3 = "no"
adSpecial4 = "no"
serviceCenterComment = "no"
serviceCenterSubscribe = "yes"
serviceCenterReprint = "no"
serviceCenterPrint = "yes"
serviceCenterClickability = "yes"
serviceCenterEmailPage = "yes"
serviceCenterEmailArticle = "no"
%>
Smokin' Supercomputing
CONVERSATION | Paracel's Jason Mollé talks to Kevin Davies about solving the IT challenges of biological analysis.
INTERVIEW BY KEVIN DAVIES
Feb. 18, 2004 |
Paracel, located a few blocks from Caltech in Pasadena, Calif., provides high-performance computing solutions to organizations big and small, including Merck, Aventis, Amersham, and Jackson Labs. Established in 1992, Paracel was acquired by Celera Genomics in 2000, and remains a wholly owned subsidiary of Celera parent corporation Applera, offering a variety of customizable systems including GeneMatcher, BlastMachine, and its new Cyclone Linux cluster. Editor-in-chief Kevin Davies met with president Jason Mollé to discuss what sets Paracel apart.
Q: Jason, you arrived at Celera prior to the acquisition of Paracel?
A: Craig Venter recruited me [in 2000]. I was tasked with being the manager of the online information business. Prior to Celera, I was president of Dialog. My background is primarily in the information side of the life sciences ... Celera acquired Paracel [later] in 2000. The original technology at Paracel was developed at TRW.
We have two parts to our business — our life science business, and our text business. Paracel really started with the latter, and that text capability allowed TRW to filter and extract intelligence from unstructured data. Still today, we support one of the largest government installations — it's a very mission-critical application.
The same challenges that exist when dealing with unstructured textual data can be transposed to biological analysis, where you get lots of repetitive data, errors, and vast quantities of data, etc. So when Paracel was spun out of TRW, the company worked with Applied Biosystems and Caltech to take that capability and create a version of it for genomic sequence analysis.
We have some unique attributes. We are a hardware and software company — we integrate and optimize the two and deliver a turnkey solution. If you look at our software accomplishments — the Paracel Transcript-Assembler, the GenomeAssembler, Pathworks, TraceTuner, Paracel BLAST, etc. — and you tie those together with very strong hardware, you get a real winning solution.
What role did Paracel play in assembling the human genome at Celera?
There were a number of parts in the sequencing and annotation pipeline where Paracel's technology was applied — in enabling the base calls coming off the sequencers to be analyzed and the filtering of those sequences through the annotation and analysis process. There were lots of pieces in the pipeline.
Is Paracel still an important part of the Celera system now that Celera has standardized on IBM?
When you look at Applera as a whole, including dealing with their back-end office administration, they made a decision on IBM, because it's more than just the early phases of analysis in research and development, but an infrastructure decision for their entire enterprise, including ERP, accounting, and so on ...
Part of the Applied Biosystems offering is the iScience portal to design, analyze, and build experiments. Within it they have a product called Panther — which classifies proteins — for which they run a lot of Hidden Markov Models. Basically, it would have brought the whole data center to a grinding halt. So they called upon Paracel to solve the problem. With the GeneMatcher, it takes a couple of days!
How was your relationship with Venter?
Craig Venter was always a fantastic character. So in 2000, Craig was sitting on top of the world, his face was on the cover of Time, BusinessWeek, he was sitting on a market cap of maybe $12 billion, he could acquire anything. The question you might ask looking back is why didn't he acquire Merck or something like that? What did he acquire? He acquired Paracel — the only substantive acquisition he made. Craig always talks about the computational challenges in biology ... He bought substance and quality here. It's a tremendously hidden story.
There were rumors that Paracel might close, but now you're profitable?
Yes, as a company, we're financially stable. We had the bioinformatics bug that bit everyone in 2000, money, no tomorrow, we had a strong parent in Applera, and we got ahead of ourselves ... We've pared the business down to where we have enough technology and resources to maintain what we currently have and what we're focused on. We've got growing revenues, a growing customer base, and a 90-percent renewal rate on maintenance contracts, which is very high in the IT space. Our job now is to turn from selling technology to selling applications. Chip design, splice variants, RNAi, we need to get to the mass market and show how we can benefit their research.
You seem particularly proud of your benchmarks.
Take BlastMachine, our Linux cluster with optimized Paracel Blast loaded onto it. It's amazing the performance we get out of that system! There's a big pharma company that was very proud of its $1-million-plus new SGI system and gave us a benchmark, expecting us to crash, and we ran the benchmark in nine hours ... They proudly pointed out that their new $1-million system ran it in six hours, but we pointed out our system only costs $50,000! We've optimized the hardware and the software — put two $50,000 systems together and it can smoke a state-of-the-art system at a top-five pharma.
GeneMatcher is the only real computational platform I know of that was specifically built for biology. Show me another chip that was specifically developed to do biological analysis. Show me a company in the marketplace today that has the biological domain expertise, the proven track record ... and the computational infrastructure. There are very few companies. In fact, I would say there's only one ... and that's Paracel.
Photo of Mollé by Gilles Mingasson