Oct 17, 2005 | Many informatics pioneers discovered the hard way that selling software licenses and fee-for-services can’t always pay the bills. Lots of companies went broke trying, a shrinking contingent struggles on, and a great many others, such as Tel Aviv-based Compugen, are working to reinvent themselves as drug discovery and development players.
Founded in 1993, long before the human genome was sequenced, Compugen was an early leader in providing predictive biology tools that combined mathematics and computers. Its initial powerful sequence search tools were followed by innovative data-mining tools (LEADS platform, 1997) and a robust wet lab (1998) to validate its in silico predictions.
The science was solid. Compugen’s clever algorithms discovered antisense RNAs are transcribed far more often than previously thought (reported in Nature Biotechnology, March 2003). It also discovered a VEGF variant protein (VEGF114) and was awarded related patents. But good science didn’t stem the mounting losses — as it hasn’t for many informatics players — and roughly a year and a half ago the company faced the unpleasant fact that its business model was stuck and unlikely to come unstuck without major change.
Last February, Compugen issued the following guidance: “Essentially all of the company’s revenues to date have been related to the licensing of platforms and tools, and the provision of services. Compugen has discontinued and/or substantially de-emphasized these activities to focus on the development and commercialization of therapeutic and diagnostic products based on the Company’s discoveries.”
It was the most direct statement of its new strategy. “Revenues in 2005 and thereafter will largely depend on royalties and other payments associated with such activities,” continued the guidance, and “Such revenues are not projected to be material during 2005.”
In May, the company announced Alex Kotzer will become president and CEO, effective in September, replacing Mor Amitai, who held the job since 1997. Kotzer is a longtime Serono executive who’s had various responsibilities including neurology and immunology product and process development, as well as worldwide manufacturing of the active ingredients for all recombinant biological drugs. Kotzer is said to be a strong manager.
Last June, Compugen announced collaborations with Biosite and Ortho-Clinical Diagnostics (a Johnson & Johnson subsidiary) in which they will use Compugen-discovered biomarkers and technology to develop diagnostics. It has a similar deal, struck late last year, with Diagnostic Products Corp. And in April, Compugen announced a multiyear “systems biology” deal with Novartis aimed at technology development rather than revenue generation.
“We’re looking to make discoveries and license them to Big Pharma. This is the business model. But [the] Novartis project is not like that. It’s not that we are seeking to gain royalties from this particular project, but we feel if in order to develop the next generation of predictive technologies for us we have to have partners with data and the guidance,” says Alon Amit, VP science and technology, commercial operations, Compugen USA.
The Ortho-Clinical deal is representative of Compugen’s revenue model. Ortho-Clinical has the right to select up to nine biomarkers from Compugen’s portfolio for development. Compugen will receive milestone payments and royalties for each diagnostic product developed. The Novartis systems biology deal is interesting and somewhat contrary to Compugen’s grand design to sell homegrown leads to partners. It’s a technology development effort.
“Historically what we have been working on with Novartis since 2001 is really just an infrastructure for genomic research. The technology at the heart of the initial cooperation was LEADS and Compugen’s modeling of the human transcriptome and proteome. It’s still being used today. But the new project is kind of different,” says Amit.
The new collaboration seeks to build models of transcription factors and their behavior. Compugen provides modeling expertise. Novartis brings data and domain expertise — critical elements enabling Compugen to develop new technologies, according to Amit.
“We’re starting small, by trying to combine data that Novartis has from different sources and see if we can come up with a better understanding of a specific type of molecular process, which in this case is transcription. Obviously there are databases today that do that. If you ask the question, what are the transcription factors and what do they bind to, you find answers in the public domain, but the answer is far from being complete or even accurate,” he says.
“The current project has a very specific goal of improving Novartis’ understanding of [transcription characteristics and behavior], and second, demonstrating the improved understanding is correct. Compugen is building the model to take experimental data and create specific answers. Those answers will be tested in the lab to see if they’re correct. If they are, from then on, you have this modeling tool you can use reliably without the need to repeat the experiments. It’s a lot cheaper than running experiments,” says Amit.
The models themselves use typical approaches — differential equations, Bayesian Network approaches, support vector machines. Compugen will own whatever models emerge and integrate what’s learned with internal discovery programs.
Clearly it’s too soon to know if Compugen’s bet will pay off, but the early signs are promising.
Amit argues the strategy is sound and that that Compugen’s wet-lab strength will help differentiate it from many in silico competitors. “One barrier to entry is that we not only create these models and fancy mathematics but we make sure the results are taken to the lab [and] actually proven,” he says.
The lab is on the same floor as the model builders and has steadily grown over the years. It is staffed by about 45 of the company’s 120 employees. “For a company our size it’s a major investment without which we think the chances that any model will actually be successful are very low,” says Amit.
Having a lab also speeds research. “Imagine working with Big Pharma overseas or in another part of the country. The back and forth that you can expect is a lot slower than if you have the biologists and mathematicians literally on the same floor discussing what to test, how to test, do the tests, and inform the models,” he says.