Shaw Lab Spinout to Commercialize ASIC-Based Molecular Dynamics System
David Shaw, founder and CEO of financial technology firm DE Shaw group and an adjunct professor in the department of biomedical informatics at Columbia University, has created a new company to commercialize technology to dramatically speed up molecular simulations.
In a keynote at Cambridge Healthtech Institute’s Bio-IT World conference in Boston this week, Shaw said that his lab at Columbia has developed a hardware system based on ASIC chips and an accompanying software program that can run a typical molecular dynamics simulation more than a thousand times faster than IBM’s BlueGene system.
Shaw said that his lab has formed a spinout firm to commercialize the technology, but did not offer further details beyond the fact that the system should be available in early 2008.
Fujitsu is manufacturing the ASICs for the hardware system, which is called Anton, after Anton van Leeuwenhoek, the inventor of the microscope, Shaw said. The software package, called Desmond, was designed for the ASIC machine but can also run on a standard compute cluster, he said.
In initial benchmarks folding a 23,558-atom dihydrofolate reductase molecule, the Gromacs package, running on a single processor, was able to generate 0.4 nanoseconds of simulated time per day, while IBM’s Blue Matter software, running on a 512-processor BlueGene system, ranged from 4.8 nanoseconds per day to 8.6 nanoseconds per day, depending on the configuration of the machine.
Desmond, running on a 512-processor cluster, reached 13.7 nanoseconds per day, and as much as 173.4 nanoseconds per day when Shaw “tweaked” the parameters of the software. On the Anton system, 512 ASICs, the standard version of Desmond reached 6,600 nanoseconds per day and the tweaked version hit 14,500 nanoseconds per day.
Shaw said that while reaching the microsecond range is promising, his goal — as well as that of many other researchers working on protein folding — is “a single, millisecond-scale simulation,” which is a time scale of great interest for biological processes such as protein folding, small-molecule binding, and protein interactions.
He noted that even though the system is much faster than Blue Gene when it comes to molecular dynamics, it is not as flexible as IBM’s system, which was designed as a multi-purpose machine. Anton can only run molecular dynamics, he said.
Shaw also cautioned that the effort “assumes that speed will allow us to do different things” than have been done with previous molecular dynamics systems. “I’m not sure that’s the case,” he said, noting that it’s still not certain whether current force fields will hold up at dramatically higher speeds. “They may have faults that can’t be seen at current time scales,” he said.
“It could just be a very fast way of getting the wrong answer,” he said.
In response to a question after his talk, Shaw said that his lab has not yet participated in the Critical Assessment of Techniques for Protein Structure Prediction experiment because the system is “not ready for that yet.” He added that he might consider taking part in CASP in the future if Desmond and Anton work as planned.
Sun Adds Life Science ISVs to On-Demand Compute Offering
Sun Microsystems is partnering with independent software vendors in the life science market to enable their tools to run on its Network.com pay-per-use high-performance computing offering, company officials told BioInform this week.
The initial version of the pay-as-you go service, which costs $1 per CPU-hour, required users to port their own applications to the on-demand platform, Rohit Valia, group marketing manager for Sun, told BioInform.
Now, he said, the company has streamlined its licensing system so that ISVs can publish their applications through the Network.com site and users can run them on Sun’s 1,000-CPU cluster with a single click.
Current life science ISV partners include SimBioSys, Q-Chem, and Biocceleration. In addition, Sun has released a suite of 13 open-source bioinformatics tools through the Network.com portal, including Blast, ClustalW, Emboss, Hmmer, Gromacs, and T-Coffee.
Sun is offering a free trial of the service to prospective users, who can enter the promotional code SGP_BIT_507_P1 to access 50 free CPU-hours on Network.com.
Sun has also expanded the offering beyond the US to 24 countries in Europe and Asia.
SGI to Bundle eXludus Grid Optimizer With Clusters for Life Science Market
SGI is partnering with workload optimization firm eXludus to include the company’s Grid Optimizer software with its Xeon clusters for the life science market.
Deepak Thakkar, SGI’s bioscience segment marketing manager, told BioInform that SGI plans to pre-install the Grid Optimizer process-scheduling software on its “standard x86 clusters” for life science customers.
Thakkar said that in internal benchmarks, the software increased the output for Blast queries by up to 60 percent without any adjustments to the software or hardware.
“It was the same query, the same data, the same system,” Thakkar said, adding that Grid Optimizer increased CPU usage from 25 percent to 75 percent.
Stephen Perrenod, vice president of sales and marketing for eXludus, said that the company’s technology is applicable to all high-performance computing applications, but said that the firm is seeing the most interest in the life science market because researchers run so many jobs in parallel that are “ideal for Grid Optimizer.”
Benoît Marchand, CEO and founder of eXludus, added that researchers in the life science market also value the price/performance ratio of their HPC systems more than some other vertical markets like engineering and financial services, which may be able to withstand the costs of inefficient CPU usage a bit better.
Expanded Servier Collaboration Brings New Technology to GenomeQuest
Officials from GenomeQuest said that a recent collaboration with French pharmaceutical firm Servier has resulted in a new software package that the company will sell as part of the latest release of its GenomeQuest sequence-search software (see Upgrades and Downloads, this issue, for details on the new product).
Servier has been a GenomeQuest (formerly Gene-IT) customer for several years, and recently asked the company to develop software to connect microarray data to qPCR data and its own proprietary reagent data sets.
“The problem had to do with trying to understand how probes from various microarrays and qPCR results lined up with various transcripts,” Michael McManus, vice president and general manager of GenomeQuest, told BioInform. “They provided all this data they had, and we included it in the system we built, and we built a piece of software that we were able to keep as part of this collaboration.”
Servier is using the software, called the GQ Gene Viewer, to view its own data in the context of information from microarray providers and public data sets. McManus said that GenomeQuest’s commercial version of the software will offer the same capability to its customers.
As part of the project, GenomeQuest created gene catalogs for human, mouse, and rat.
GenoLogics Eyes Translational Medicine with New Omix LIMS
GenoLogics has launched Omix, a cross-platform laboratory information management system that the company is targeting to the translational medicine market.
GenoLogics CEO Michael Ball said the system integrates genomics, proteomics, and metabolomics data so that it can be shared with information from clinical systems.
Drug-discovery firms are currently “wrestling with linking clinical data with omics data” to discover biomarkers, Ball said. Omix provides a “common infrastructure” for companies to enable that linkage from a single source, he said.
Ball said that the technology behind Omix was previously embedded in the company’s Proteus and Geneus LIMS, which were developed to manage proteomics and genomics data, respectively. Now, he said, GenoLogics has built that underlying technology into a standalone product that can be configured to work with customers’ in-house experimental platforms.
In addition, Proteus and Geneus are available as application-specific modules that sit on top of the Omix framework.
Several of the company’s current customers, including Windber Research Institute, Biogen Idec, the University of Virginia, Penn State College of Medicine and Penn State Hershey Medical Center, Scripps Florida, and the University of Glasgow, are using Omix for translational medicine applications, the company said.
Ball said that Omix is integrated “out of the box” with 30 different proteomics and genomics experimental platforms.
James DeGreef, vice president of product strategy, said that GenoLogics plans to develop additional application-specific modules to plug into Omix in the future. Likely areas include metabolomics and RNA interference, he said.
CLC Bio to Launch RNA, MLST Software This Month
CLC Bio plans to release several new software packages for RNA analysis and multilocus sequence typing by the end of the month, Jan Lomholdt, vice president of global sales and marketing, told BioInform.
The upcoming CLC RNA Workbench will include algorithms for predicting RNA secondary structure, a graphical interface for editing RNA secondary structures and sequences, and analytical tools for motif finding.
The MLST module is designed to quickly and automatically type bacteria and yeast isolates, Lomholdt said. For each gene, the experimental data is compared to a database of known alleles and a sequence type is assigned by combining information from all genes.
SeiraD to Ship Gentellect Data-Integration Software in August
SeiraD, a startup developing next-generation sequencing technology, plans to release a new software platform for data integration in August, company officials told BioInform.
The firm was awarded a grant from the US Air Force to develop the software in collaboration with software firm Ananomouse.
The software, called Gentellect, integrates genomic, protein, and pathway data.
Joe Gatewood, president of SeiraD, said the Air Force and two undisclosed pharmaceutical firms are beta-testing the software.
Bioinformatics.Org Opens Core Facility for R&D, Education
The Bioinformatics Organization, a non-profit institute that supports open-source bioinformatics development, has opened a core facility to offer research, development, and educational services, Jeff Bizarro, founder and chairman of Bioinformatics.Org, told BioInform this week.
The fee-for-service facility expands the commercial offerings of the organization, which began offering fee-based bioinformatics courses last year [BioInform 06-02-06].
The facility will offer bioinformatics services and support for several research platforms, including flow cytometry, high-content screening, microarray data analysis, SNP genotyping, sequence analysis, mass spectrometry, PK/PD modeling, systems biology, cheminformatics, and literature mining.
Early clients for Bioinformatics.Org’s educational courses include Amgen, Sigma-Aldrich, Wyeth, Transgene, the National Institutes of Health, the US Department of Agriculture, Max Planck Institute, and Baylor College of Medicine.
Bizarro said that the organization has also begun its first R&D project with the Midwest Research Institute.