Next year, the company that designed the National Cancer Institutes microarray database and analysis software is planning to offer a product to the commercial market.
The company, SRA International, developed MicroArray DataBase, or MADB, for the NCI, and the system has nearly 500 users and holds 10,000 microarray images.
The US Centers for Disease Control and Prevention, as well as the Netherlands Cancer Institute (NKI) and the Genome Institute of Singapore also use versions of the software tailored for their needs.
Since these systems have been so popular, and offer the unique ability to handle large-scale clinical datasets, SRA realized it had something that the commercial market might need. So for the past 15 months, SRA has sought to develop this product, which it calls BioSilk.
Recently, SRA decided to spin out Tapestry Technology as a stand-alone company to develop and commercialize the BioSilk software for other labs. Gary Nelson, a high-ranking SRA executive, decided to leave his post to spearhead the effort.
In an industry crowded with microarray analysis packages, from GeneSpring and Rosetta Resolver to newcomers such as GeneSight, this decision took verve.
But Nelson sees the BioSilk product as a second-generation array analysis package that could supercede current offerings. Sure, it offers the usual cluster analysis tools such as agglomerative algorithms, k-means, consensus clusters; and offers its own genetic algorithms. But the important difference, Nelson stressed, is the capability to handle thousands of arrays and link array information to other data.
We see the second-generation microarray analysis software as merging of the microarray data with genomic and clinical data so you can test scientific hypotheses rather than just looking at patterns, he said.
The software will be able to cluster not only expression patterns in different ways, but also link expression to literature searches and to the samples clinical characteristics, such as its disease state.
Other features that BioSilk plans to offer which are not in currently available packages include Bayesian statistical techniques and machine learning technology, trainable algorithms of which neural nets are a prime example.
The program, which is designed as a server-based solution, is targeted for release in mid-2002 according to Nelson.
But before BioSilk gets to the laboratory desktop, Tapestry still has a lot of weaving to do. The company is seeking financing partners to further commercialize the technology, and has not yet decided on a hardware platform.
Financing would enable us to complete development and get it to market, Nelson said.