A number of companies promise to predict a drug's toxicity based on gene expression profiles. But SimuGen, a computational biology startup based in Cambridge, UK, believes it can do this better, cheaper, and faster.
"A lot of potential competitors seem to be focusing on the 'better' aspect, offering extensive microarray services," said Quin Wills, a company founder. "What we feel the market really needs is a quick screen, something that you can run in your lab … rather than sending off your molecules to some service and getting some results two months later," he said.
SimuGen is hoping to launch its first product, a kit that will allow researchers to assess liver toxicity, early next year. The kit, whose manufacture will be outsourced, will consist of a cell line and chemistry to do quantitative PCR on "a handful" of marker genes chosen by SimuGen, and software to connect to the company's server. Priced at about £2,500 ($4,500), the kit will allow users to test up to 10 substances at eight different concentrations.
The company is "always willing to tweak a kit to run another cell type if that is what a customer may prefer," Wills said. However, he said, for reasons of reproducibility, SimuGen will start with one "well-studied and accepted immortalized cell line" for the first kit.
"Microarrays can only point you in the right direction. The quality of the data is not particularly reproducible, they are all very expensive, [and] the average person doesn't know really how to use them."
SimuGen will provide the data analysis: The company uses various clustering and machine-learning methods to compare PCR data provided by its customers to an in-house database of known gene expression profiles of toxic reference drugs. "No processing of data in any way is expected from the customer," Wills said. "Our server will process the data, ensure that QC standards have been met, and then feed the results back to the customer."
That database is still in the making. Based on publicly available and in-house expression data and the literature, Wills said that SimuGen scientists have come up with "just a handful" of candidate markers for its first kit, which they are now having tested on a number of cell lines for a range of drugs with known toxicity. These drugs were chosen because they were "prototypical enough to give clear results, but not too narrow in effect that the system is overtrained to recognize only particular pathologies," Wills said.
The company is also exploring an option that will allow customers to add their own reference molecules to the database. "One option is to allow any known customer to submit data together with basic descriptions such as chemical structure," Wills said. "This will be treated and presented as third-party data, though SimuGen may maintain a curation process whereby we add well-annotated and supported data to our main reference databank for customer use."
Wills would not divulge the nature of the clustering methods the company uses to decide whether a molecule is toxic or not, but the emphasis is on "accurate" and "simple to understand," he said.
"There are plenty [of] approaches out there, but most [are] completely beyond the understanding of the average biologist," he said. "We have one or two techniques that are proprietary that can keep the clustering really nice and tidy and give you some very easy-to-understand results." The output is "a very simple ranking system of drugs based on their toxic profiles at different doses together with visual outputs that will allow one to assess toxic parameters like potency at a glance."
In contrast to competitors such as Gene Logic and Iconix, which maintain databases of microarray-based gene expression profiles correlated with toxicity, SimuGen believes that a small list of well-curated marker genes is more useful for toxicologists looking to make a quick decision about whether to proceed with the development of a drug. Another potential competitor Wills cited is CuraGen, which offers an array-based predictive toxicology service based on expression profiles of marker genes it has identified from rat hepatotoxicity experiments. But the reliance of these firms on microarrays sets SimuGen apart in the marketplace, according to Wills.
"Microarrays can only point you in the right direction," he said. "The quality of the data is not particularly reproducible, they are all very expensive, [and] the average person doesn't know really how to use them."
SimuGen does rely on microarrays for the development of its markers, however, "and we will use microarrays on our drug libraries whenever we find we need to beef up data to screen for markers," Wills said.
At the moment, the company is busy finishing its proof-of-concept work before it gears up to launch its first product in early 2006. Said Wills, "The question is, 'Is this robust enough, will this work as a kit? Will a toxicologist like this?'"
To market the kit to pharmaceutical and biotechnology companies, SimuGen seeks to partner with at least one commercial toxicology service, but will also maintain some direct sales to stay in close touch with customers and "keep a pulse on how the kit should evolve," Wills said.
The company is also looking to close a financing round by the end of the year that will bring its total cash position to about £1 million, a goal that Wills is optimistic the company will meet. SimuGen's early-stage funding — an undisclosed sum from angel investors, venture capital companies, and matching grants from UK government bodies — will allow it to finish its proof-of-concept studies, he said.
At present, SimuGen's staff consists of the company's four founders — all with strong ties to Cambridge University and affiliated institutes — as well as a chairman and a financial director. The company is currently looking to hire a chief operations officer as well as a computational biologist with experience in microarray analysis who will jointly lead the team scientifically with Wills.
"To be excellent at computational biology as a company, you need to have a very broad spectrum of skills, [and] that's what we have managed to put together," said Wills. He himself is an MD and human geneticist and has been working as a computational biologist at the university. The other founders bring their expertise in business development, machine learning, and industry project development to the table.
— Julia Karow ([email protected])