India’s only academic spin-off in bioinformatics celebrated its first birthday with the launch of the country’s first commercial bioinformatics product: image classification software to accelerate high-throughput crystallography in protein structure determination.
Strand Genomics, founded by computer scientists at Bangalore’s Indian Institute of Science (IIS), said the software is the first ever to automate the classification of images of the crystal growth process.
The software, called Sphatika, will help select the most promising images of potential crystals that can be subsequently put through the X-ray crystallography process.
High-throughput crystallography generates up to a million images of the crystal growth process a day that need to be scanned to reject non-crystal entities and to identify potential crystals for further study.
Image scanning is currently a manual process. Crystallographers, typically, have to scan tens of thousands of images to pick potential crystal images.
“It’s a tremendous waste of expensive and hard-to-find crystallographer time,” said Sreenivas Devidas, vice president of business development strategy at Strand Genomics. “The probability of finding crystals during the process of crystal growth is extremely small and that makes the task all the more difficult.”
Protein crystals are grown under various combinations of parameters like acidity and temperature to find the best for crystallization. While automated robots alter the parameters, images are taken to determine whether crystals are forming. The complexity and diverse nature of these images has made it hard to automate image classification.
Strand’s software will classify these images, eliminating or reducing the need for human eyes to pick out promising crystals for 3D structure analysis. The elucidation of structures of tens of thousands of proteins has emerged as a major goal in post-genome sequence biology.
At least four companies are involved in high-throughput crystallography for protein structure determination — Structural Genomix and Syrrx in California, Integrated Proteomics in Canada, and Astek Technology in the UK. They’ve set ambitious targets. San Diego-based Structural Genomix, for instance, plans to unravel at least 5,000 protein structures(?) over the next five years.
“The new software can process an image and classify it in five seconds. Running on a Linux cluster, it could process up to a million images a day,” said Ramesh Hariharan, co-founder of Strand Genomics and chief architect of the product.
“The classifier improves productivity of a crucial step in the process of determining protein structure via X-ray crystallography,” said Hariharan, who is an associate professor of computer science at the IIS. “It could turn out to be a critical step in getting ahead in the game,” Hariharan said.
Strand Genomics said it is currently finalizing an exclusive deal with a leading player in high-throughput crystallography.
Four computer scientists at the IIS teamed up to create Strand Genomics in November 2000. The approval by IIS to allow working academics to launch a commercial venture was hailed as a landmark event in India.
Traditionally, Indian universities have discouraged researchers from setting up businesses and academics have been little inclined to commercialize innovations.
Strand Genomics is also working toward a mid-2002 release of a suite of products called Oyster designed to boost productivity in the drug discovery process. The company is trying to expand its scientific workforce from 25 to at least 50 in the next four to six months.
— Ganapati Mudur, New Delhi, India