Biosystemix serves as further proof that good bioinformatics software doesn’t go away — it just resurfaces in new (and usually smaller) surroundings. The Kingston, Ontario-based firm — founded by two former Molecular Mining staffers — is the latest example in a series of new businesses that have sprouted from the remains of defunct bioinformatics shops. Bolstered by proven technology and established contacts, the fledgling firm has found that success in the bioinformatics market is within reach for those willing to risk a fresh start.
“The field is not now in the period of over-optimism and over-confidence that we saw in 1999 to 2000,” said Roland Somogyi, president of Biosystemix. Nevertheless, he noted, “Larry [Greller, Biosystemix co-founder and CSO] and I felt that this is a good time to form a smaller company that delivers very focused services, very customized to particular customers, and build from there, and have this grow organically.”
Biosystemix has licensed two key algorithms from Parteq Innovations, the technology commercialization agency for Queen’s University in Kingston, which has held the rights to Molecular Mining’s technology assets since the company’s closure in early 2003. In addition to the two algorithms — SLAM (sub-linear association mining) and IBIS (integrated Bayesian inference system) — Greller and Somogyi have developed a number of new predictive modeling methods to round out their technology toolkit and address the data integration and interpretation challenges of systems biology.
The company is sticking to a services model, reviving one aspect of Molecular Mining’s former hybrid approach. Another Kingston-based startup, Predictive Patterns, resuscitated the software sales component of Molecular Mining’s offering last year by acquiring rights to sell the company’s GeneLinker Gold and GeneLinker Platinum gene expression analysis products [BioInform 06-20-03].
Splitting the Difference
This bifurcation of a defunct company’s technology assets into two new entities — one following the services path, while the other focuses on software sales — is emerging as a successful pattern in the bioinformatics sector as it regroups following the tough market conditions of the past few years. Physiome Sciences’ technology assets have been split in similar fashion by two new startups — the BioAnalytics Group and BioSoftware Systems — with the former specializing in collaborative research partnerships, while the latter addresses shrinkwrapped software sales [BioInform 07-05-04]. So far, the model appears to work. The two firms announced last week that they have jointly signed a research agreement with Roche [see briefs, p. 8, for further details.]
This type of setup is also working out well for Biosystemix. In late June, the company signed a collaborative partnership with S2K, a consortium of Canadian researchers funded by a Can$15 million ($11.4 million) grant from Genome Canada to study functional genomics, pharmacogenomics, and proteomics associated with several disease areas. The company is actually resuming an agreement that Molecular Mining had with the consortium. S2K has also signed a deal with Predictive Patterns to continue using the GeneLinker Platinum software that it had originally licensed from Molecular Mining.
Its Molecular Mining connections have helped the young firm establish trust among its potential customer base, Somogyi said, “But we’ll try to make new friends, too, of course.”
Forward Modeling, Reverse Engineering
The S2K partnership is typical of the kinds of collaborations Biosystemix is hoping to develop. While the S2K consortium includes a number of bioinformatics experts who conduct “primary analysis” on data collected from a range of experimental platforms, Abdelkader Yachou, program manager for S2K, said that the company was called in for its expertise in “higher-level” analysis. “They’ll be building bioinformatics models that will allow us to predict susceptibility and outcomes” based on the experimental data that the consortium’s researchers collect, he said.
The S2K project is focusing on three major immune dysfunction areas: infectious disease (specifically HIV, hepatitis C virus, and SARS); transplant rejection; and rheumatoid arthritis. Somogyi said that Biosystemix will identify genes and markers using first-level bioinformatics analysis, “but it’s not just about identifying the markers — it’s about identifying the predictive models in the applicable context,” he said. Biosystemix will apply a combination of data mining and predictive modeling technology to derive a set of markers in tandem with the nonlinear and combinatorial relationships that guide their behavioral interaction. “We will find these markers for S2K, but these markers are really parts of predictive models,” Somogyi said. “It’s not just, ‘Here’s a list of the five top genes.’ What S2K really needs is a predictive model that includes those five genes — and there may be alternative predictors comprising 50 genes that in various combinations of five or 10 could do the job. So the focus isn’t on the single gene any more; it’s how to find which combination of genes in which context is predictive.”
Greller said that the company addresses research problems from two directions — a “forward-modeling” approach, which builds simulation models based on information from the literature and other sources, and a “data-driven” method, which derives relationships based on patterns in large experimental data sets in order to “reverse engineer” regulatory pathways and networks. Depending on the customer and the research question, Greller said, Biosystemix could apply several combinations of the two approaches.
In addition to the S2K agreement, Biosystemix has similar collaborations with Canada’s Queens University in the area of follicular lymphoma; with Michigan State University to predict the toxic response pathways of steroidal drugs; and with the University of California, San Francisco, to predict interferon-ß response in multiple sclerosis patients. Somogyi said the company just signed an agreement with the Immune Tolerance Network to derive “complex patient signatures” associated with autoimmune disease. The company is also in discussions with two mid-sized biotech firms and several large pharmas regarding possible collaborative partnerships.
Biosystemix faces some competition from biosimulation companies like Gene Network Sciences, Entelos, and the BioAnalytics Group, as well as systems biology research firms like Beyond Genomics, but Greller said that his company’s focus on providing bespoke predictive models sets it apart from others in the field. “Where we see the most competition,” Greller said, “is from in-house groups” in pharmaceutical and biotech companies. The goal in these situations, he said is to “convert that competition to cooperativity, where their expertise complements ours.”
Greller added that the company’s small size and services-based business model provide a number of advantages over Molecular Mining’s approach. For one thing, he said, the revenue-driven firm is “unencumbered by investor goals, which is okay at this stage for us. It gives us a lot of freedom to operate.” Another advantage, he noted, is that “we do all this in research code, so at any time our explorations can be fully baked or [a] quarter baked — that doesn’t slow us down. If your business requires you to convert some of your new technologies into a shrink-wrapped software product, it becomes much harder — with great proportions of your efforts devoted to software engineering and development,” he said.
Biosystemix is targeting what is presently a niche segment of a slowly recovering market — a strategy that may require the company to remain small for the time being. But that’s just fine with the company’s founders, Greller said. “It became clear to [us] that if we keep pursuing this without a business model that demands a large payroll, and a large team of technical professionals, then it becomes financially and operationally feasible to deliver and make this work.”