By Tiffany Fox, (858) 246-0353, firstname.lastname@example.org
San Diego, Calif., Jan. 6, 2013 — If the computational “cure” for cancer is out there, Rommie Amaro and her team at the University of California, San Diego intend to find it.
They know, however, that they can’t do it alone. That’s why they’ve developed a computer-aided drug design simulation platform that anyone in the research community can use, at unprecedented speeds, to find promising drugs for the most pernicious forms of cancer.
The NVIDIA Foundation — the philanthropic arm of the California-based graphics processing units (GPUs) manufacturer — is so convinced that Amaro is onto something that awarded her $200,000 last month to develop the platform as part of its Compute the Cure initiative.
Amaro’s automated computational workflows make ample use of GPU-enhanced technology to understand the molecular dynamics of a protein known as p53, which, when mutated, is implicated in more than half of all human cancers.
“What this means is that anybody who has a GPU card in their computer can run the simulations they need and make great discoveries in the fight against cancer,” says Amaro, who is an Assistant Professor of Chemistry and Biochemistry and is affiliated with the UC San Diego division of the California Institute for Telecommunications and Information Technology, also known as the Qualcomm Institute (QI).
Amaro’s team includes Associate Research Professor Ross Walker, who developed the molecular dynamics code for the platform, and Deputy Coordinator for Research Ilkay Altintas, who designed the BioKepler workflow framework on which the platform is based. Both are affiliated with the San Diego Supercomputer Center, and Altintas is also affiliated with QI.
“We have all these single 3D snapshots of the p53 protein from X-ray crystallography experiments, but we know that these proteins structures are not static structures — there’s all this intricate machinery that’s constantly moving and reacting and carrying out processes required to sustain life,” Amaro explains. “With molecular dynamics simulations, we use computers and pretty simple physical equations to bring static, experimental molecular structures to life and then, find drugs that will target them.”
Amaro says that predicting the atomic-level motion of a drug target is very important, especially for a target like p53. It’s the protein that controls cell suicide, but when it mutates the resulting cancer cells don’t have the capacity to know when to die. These damaged cells then keep proliferating.
“The idea we’re all working on is to develop a chemotherapy that would cause cells to reactivate p53 and die, but we can only work on only so many anti-cancer drug targets at a time – maybe three to five,” she continues. “If we can make these tools available to use these GPU-enabled technologies, the power of these technologies can scale much more broadly.”
Amaro notes that since her lab started exploring the structure and simulating the dynamics of p53 with GPU-enabled computers, they’ve discovered “a new druggable ‘pocket’ that was not present in the crystal structures, but reveals itself in these computational simulations.
“This pocket allows us to not only understand how these particular compounds work with pretty good certainty, we have also used this new pocket to discover a whole host of compounds that would will hopefully be able to reactivate p53 in a large number of cancers. With this approach, it’s possible that we can develop one drug for a broad spectrum of cancers, which is exciting.”
Amaro is working in collaboration with NVIDIA and the National Institutes for Health to pursue experimental trials to determine the efficacy of these compounds in animals and then humans, a process that she calls “the slow part” of the drug discovery process.
But the computational steps to finding a cure for cancer are made “orders-of-magnitude” faster by the GPU cards, says Amaro, “and we’re still getting faster with every cycle.”
“It turns out that GPU cards, the same types of silicon chips used to develop video games, can actually be super useful for scientific computing as well,” she adds. “Previously when we were limited to CPUs (central processing units), it might take us weeks to months to run those simulations and find new 'druggable' pockets. What took us a week or longer before we can now do overnight. For p53 reactivation, I can say that out of approximately 120 compounds that we’ve experimentally tested, we have a viable pool of about 16 to 17 compounds, and which took us only two to three weeks to discover. The NVIDIA GPU cards are like having a supercomputer on your desktop machine.”
Amaro, who says she is “passionate about enabling others to use our technologies,” notes that her automated workflows (which she calls “computational recipes”) are cheap to run, too. “A lot of people already have these cards in their normal computers, and they can be used by people who don’t have the same level of laboratory experience and training that we do.”
The team is leveraging key partnerships to help with the distribution of the workflows, including its collaboration with the National Biomedical Computational Resource (NBCR), which is based at UC San Diego and led by Amaro, as well as the Teach-Discover-Treat initiative, a grassroots effort to provide high-quality computational chemistry tutorials that impact education and drug discovery for neglected diseases.
Amaro says another collaboration she’s particularly enthusiastic about is a partnership with the National Center for Microscopy and Imaging Research (NCMIR), which is also based at UC San Diego and is affiliated with the Qualcomm Institute.
“We’re always looking to improve our computational techniques, and we know that, in reality, these drug targets are actually in a very elaborate biological milieu,” notes Amaro. “What we’re trying to do in collaboration with NBCR and NCMIR is build new computational technologies that will allow us to actually simulate these targets in more realistic in vivo environments.
“NCMIR can obtain beautiful maps that show us where cells and bits of the cellular machinery are located. We’re now trying to figure out how we can take images of these ‘nano-neighborhoods’ and use them to construct more elaborate in silico models, which we can then bring to to life with simulation. A better understanding of the actual in vivo environment of a drug target can show us how molecules are systemically distributed and what they interact with. Achieving more realistic models will give us a better chance to predict and engineer small molecule compounds that have the potential to have big impact in human health.”
Amaro adds that her workflows are so accessible that Canyon Crest High School student Eric Chen, who worked in her lab in Urey Hall last year, was able to use them for drug discovery. Chen later went on to win the Grand Prizes in this year’s Google Science Fair and the Siemens Competition.
“There are so many different, possibly viable anti-cancer targets, and historically we’ve been limited as to how many of them we can work on,” she remarks. “These technologies are definitely the way of the future — they have the capacity to make experimental efforts more efficient and effective. Our algorithms are getting better at being predictive and at the same time our computational capabilities are enabling entirely new paradigms of biomedical research to be developed.
“If we can put these technologies in the hands of others, it becomes something that scales much more effectively,” she adds. “With all the different kinds of cancers that we are faced with, we need all the help we can get to win this war.”
Tiffany Fox, (858) 246-0353, email@example.com