London-based startup Matrix Advanced Solutions is developing a computational drug-design platform to address the limitations of current approaches both computational and experimental.
The company's technology, called CADDIS (Computer-Assisted Drug Discovery), uses a genetic algorithm to optimize lead molecules based on the concept of "fitness" as defined by drug-like qualities. The first version of the software, published in 2002 [Kamphausen et al., "Genetic algorithm for the design of molecules with desired properties." J Comput Aided Mol Des. 2002 Aug-Sep;16(8-9):551-67], was able to optimize molecules based on single parameters toxicity, selectivity, metabolic stability, or the like. More recently, the company has improved the approach to optimize multiple parameters simultaneously an improvement that is expected to generate compounds with a better chance of surviving clinical trials.
The first results of the new method as applied to the design of thrombin inhibitors were published in a recent issue of the Proceedings of the National Academy of Sciences [Riester et al., "Thrombin inhibitors identified by computer-assisted multiparameter design." Proc Natl Acad Sci USA. 2005 Jun 14;102(24):8597-602].
Martin Wiesenfeldt, CSO of Matrix Advanced Solutions, likened the benefits of the multiparameter approach to a travel agent seeking the best resort for a customer who wants to both hike and swim. "You can take a map and look for the highest mountains. Starting from there, you go downhill until you find the first lake. But I don't think this is a good way to spend your holidays with respect to swimming. On the other hand, if you start at the beach and climb the first small hill, it won't be ideal either," he said. "So taking all relevant criteria right from the beginning takes you to completely different places where both requirements or criteria are fulfilled."
"We usually end up with something close to the best compromise with respect to several parameters."
Andreas Schweinhorst, a group leader at the Institute of Microbiology and Genetics at the University of Göttingen and co-author on the PNAS paper, said that other lead-optimization approaches, such as high-throughput screening and virtual screening, often produce molecules that are "stuck somewhere in an early and partial solution." With CADDIS, however, "We usually end up with something close to the best compromise with respect to several parameters," he said.
The thrombin inhibitors discussed in the paper "are close to the best compromise that can be constructed into molecules, and therefore they should have a higher probability to survive in clinical phases," Schweinhorst said.
In the paper, the authors described the method, which is based on "design cycles" that combine combinatorial synthesis, experimental characterization of the compounds, and computational processing of the data to derive a set of compounds to be synthesized in the next cycle.
Using this iterative approach, the authors started with a set of 170 randomly chosen compounds and performed eight design cycles to identify a set of lead compounds that "show extremely favorable properties concerning thrombin inhibition, selectivity, metabolic stability, (low) serum protein binding, (low) toxicity, activity in in vitro coagulation assays, and (slow) elimination from the bloodstream, making them attractive candidates for further development," the authors wrote. Fewer than 1,000 molecules were synthesized over the course of the eight cycles.
According to Matrix Advanced Solutions, the multiparameter approach was able to identify promising lead candidates that would have been overlooked by high-throughput screening.
Using Wiesenfeldt's metaphor of the travel agent, Sion Balass, CEO of the company, said that high-throughput screening is the equivalent of "working from the Himalayas the derivatives of the best compound and you will never end in a place where you can swim and climb mountains."
Balass said that the approach is also an improvement over virtual screening, which has "a very high inefficiency because it's simulating behavior." CADDIS, on the other hand, uses the relationship between the structure and behavior to predict optimal structures. "It is these structures that we then synthesize and we test them in the laboratory. So there's no estimation, there's no simulation here. It's real numbers, and you get the numerical values of the activities for each one of these suggested compounds, and you can see its behavior."
Another drawback of traditional lead optimization approaches, he said, "is that you don't learn from mistakes hundreds of thousands of compounds, and nobody actually learns why this one is not active, why it's not working." The company's approach is "very efficient because we actually use all the information," he said.
Balass said that the company plans to move its thrombin inhibitors through preclinical and clinical trials, and that it is also using the approach to develop antibiotics and tumor therapeutics.
The company, founded two years ago, is pursuing a two-pronged business model based on drug-design services and in-house development. "The idea is to create a company that will have its own pipeline but at the same time will also be generating revenue," Balass said.
Matrix Advanced Solutions is "in discussions" with several biotech and pharma firms, Balass said, but he declined to identify any potential customers.
Bernadette Toner ([email protected])