It’s all about the algorithms, according to Ming Li, CEO of Waterloo, Ontario-based Bioinformatics Solutions.
Unlike many bioinformatics companies, who “concentrate on putting a good user interface on free software,” Li said his company is focused on “developing proprietary software based on advanced algorithm designs to provide real solutions to bioinformatics problems.”
The eight-employee company had its roots in the bioinformatics research group at the University of Waterloo. Armed with around $500,000 in venture capital, BSI launched officially last October with the hire of its first full-time employee.
The company’s newest product, PatternHunter, is a homology search algorithm that is thousands of times faster than Blast at the same sensitivity, according to Li. While most approaches to speeding up homology searches rely on tweaking or parallelizing the original Blast algorithm, Li and his colleagues started from scratch with their approach.
The PatternHunter algorithm, which they describe in an upcoming Bioinformatics article, is based on a novel approach to seeding — the process of finding short “seed” matches which are then extended into longer alignments. While Blast looks for seed groupings of consecutive letters, PattternHunter uses groupings of non-consecutive letters, which increases the algorithm’s sensitivity, according to BSI.
The Mouse Genome Project is currently using PatternHunter to compare the mouse and human genomes, Li said, a process that would take 19 years using Blast. “At the same sensitivity, PatternHunter takes 20 days on a single PC.”
Li said that BSI is still working on adding protein sequence comparison capability to PatternHunter, a feature that should be available in a few weeks.
In addition to PatternHunter, BSI also offers Prospect, a structure prediction program based on threading techniques; Copia, which finds consensus patterns in proteins; and Hyperclean, which computes phylogenies of species.
BSI is also partnering with Caprion Pharmaceuticals to develop new algorithms for mass spec analysis. Caprion has around 20 mass spectrometers, but currently available analysis software “doesn’t meet their needs,” Li said. Mascot and other programs that Caprion has tried aren’t flexible and have other limitations. “For example, no software does de novo sequencing well,” he said.
It’s expected that the resulting mass spec software, scheduled for completion in four months, will be co-marketed by Caprion and BSI. The software will serve as one component of an “open proteomics environment” that BSI is building. A key aspect of the environment will be a graphical programming language that will allow users “to write programs without writing programs” — simply by dragging icons and connecting lines to compose computable tasks.
A prototype of the environment is ready, but Li said the company is first focusing on getting the individual software components to market. BSI is currently “self-sufficient,” Li said, but may seek additional funding before expanding the company into the larger operation that will be necessary to support the proteomics suite. Li expects this to take about a year.