Locus Pharmaceuticals said last week that it has been granted “broad patent coverage” for its computational method for predicting binding sites on protein surfaces. The patent will be the first for the Blue Bell, Pa.-based drug design firm, as well as a relative rarity in the field of computer-assisted molecular design: Most firms — if they file for patents at all — set their IP sights on novel molecules discovered using their tools, but not the technology itself.
Locus’ decision to patent “is driven by our belief that we have to get the technology out there so that people can see it and understand scientifically that it’s credible, new, novel, and actually is different,” said Jeff Wiseman, VP of technology and informatics at Locus. “We have to teach the world that we are doing something new,” he said.
As in other computational fields, the value of patenting software-related inventions in the life sciences is a matter of much debate. After all, once a patent is issued, the guts of the technology are on public view, and it’s fairly simple for a competing firm to develop a slightly different algorithm that performs the same task while evading infringement. For this reason, many software firms eschew patenting in favor of keeping their technology a trade secret (see related article, p. 7). But, as Wiseman said, making the technology public is exactly what the company wanted to accomplish in this case.
The US Patent and Trademark Office has not yet issued the patent itself, but it has released a notice of allowance, meaning that the prosecution phase is complete, and the examiner has declared the technology to be novel and patentable. According to Yuriy Stercho, director of legal affairs and secretary at Locus, the patent should issue in two to four months.
The allowed patent application is the first of several that Locus has submitted for review. Stercho said that the company’s patenting strategy positions it “on the leading edge” of the computational drug design sector, where patent protection for software is still considered “unusual.” However, he added, the resistance to patenting “is going to change, and is in the process of changing. Most of our competitors probably have something in the pipeline.”
But not all in silico drug design firms are rushing to patent their tools. Burt Wuurman, CEO of De Novo Pharmaceuticals, told BioInform that De Novo has not yet filed any patents for its molecular design algorithms, “but we are constantly reviewing that strategy, so it’s something that might change in the future.” De Novo’s strategy of keeping its methodology a trade secret has served it well so far, Wuurman said. He added that the company has carefully documented the history of its technology development process “So that if it’s challenged, we can say, ‘Look, there’s prior art dating back to there and there.’”
While many technology firms build large patent portfolios in order to coerce licensing fees from competitors, Wiseman said that Locus views IP as more of a means to attract collaborators, even among those firms that may at first glance appear to be rivals. The company has been in partnership discussions with other computational drug design firms, he said, “but it’s difficult because people take an attitude that we’re competing, and that colors your ability to have those kinds of conversations.”
Whether potential partners will view the upcoming patent as proof that Locus is not a direct competitor — or as a threat — remains to be seen, however. “That’s the big question,” Wiseman said. “That’s the gamble for putting your methods out there.”
Delivering on the Dream?
Locus is one of several companies promising to overcome the limitations of combinatorial chemistry and high-throughput screening — namely, cost, time, and inaccuracy — via computational approaches. Genomics has provided pharmaceutical firms with an avalanche of new targets, Wiseman said, but they are still “dependent on chemistry that was developed for old protein targets.” While around 100 million compounds currently exist physically for high-throughput screening, Wiseman estimated, the fragment-based computational approach that Locus has developed provides access to 10<sup>15</sup> possible compounds that can be screened in silico against any high-resolution 3D structure.
Wiseman acknowledged that the industry has “dreamed about being able to computationally design drugs accurately for twenty years, and nobody’s been able to do it.” What sets Locus’s approach apart from previous computational methods, he said, is that it is able to speed up the process by a factor of 10,000 by “totally turning around the way you look at thermodynamics and calculate binding energies.”
The “slow method,” Wiseman explained, calculates the free energy required for binding a molecule in a particular position on the protein surface — not a complicated task for a single calculation, but if you wanted to do this for the thousands of possible positions at each binding site, across the whole protein surface, and for a thousand or more fragments, it would take thousands of years of computer time.
Locus reverses this process by setting the energy level first, and then finding all the possible combinations of fragments and positions across the protein that result in that energy. “Doing it that way is thousands of times more efficient than searching across the protein surface, and you also get the binding energy out of the efficient search process. So you solve both of those problems at one time,” Wiseman said.
So far, the company has used the approach to design three molecules for its in-house programs that Wiseman said should be in pre-clinical studies by 2005. Locus also has three early-stage projects under way with partners, Wiseman said.
Locus is not the only firm seeing in silico drug design success. Last week, Predix Pharmaceuticals announced that it had begun Phase I clinical trials for a compound it claims to be the “first drug candidate discovered based on computer-generated GPCR models and optimized with integrated computational-medicinal chemistry.” The compound, a dual-action serotonin 1A (5HT1A) receptor agonist/sigma-1 antagonist intended to treat anxiety, attention deficit hyperactivity, and other neuropsychiatric disorders, took less than two years to move from in silico discovery, through optimization, and into clinical trials, Predix said.
If — after twenty years of trying and falling short — computational approaches do begin paying off this time around, Locus and other computational drug design firms may find a sharp increase in demand for their technologies, and Locus’s early lead in building its IP position may give it a crucial competitive edge.