This article has been updated to note that the current grant is part of a larger $1.7 million three-year NBIB grant for StochSS' development and to clarify the grant goals.
NEW YORK (GenomeWeb) – Researchers from the University of California, Santa Barbara have received just over $540,000 in grant funding from the National Institutes of Health to continue developing the Stochastic Simulation as a Service (StochSS) platform, a cloud-based system that provides computational resources for creating stochastic models of cellular systems.
According to its abstract, the grant from the NIH's National Institute of Biomedical Imaging and Bioengineering — its part of a larger $1.7 million grant that the group has received from NBIB for StochSS — will support ongoing efforts to develop new algorithms for simulating "fully- adaptive, hybrid solvers for stiff (and nonstiff) well-mixed systems; parameter estimation for discrete stochastic systems; and simulation of spatial stochastic systems." The funds will also support continued development of a computing environment for the aforementioned modeling software that runs on both local and cloud-based infrastructure.
The StochSS project is co-led by Linda Petzold, a UCSB professor of mechanical and environmental engineering and computer science, and Chandra Krintz, a professor in UCSB's department of computer science. Petzold's laboratory has spent the last 15 years developing algorithms and software for generating discrete stochastic simulations of biochemical systems, and through this project they hope to make those tools more broadly available in software that is simple to access and use.
"Much modeling in biological sciences is done by ordinary differential equations (ODEs)," which work well for molecular interactions in standard chemical reactions but are not as effective for looking at biochemical interactions in cells, she explained to BioInform this week. That's because in a cellular system, "there are many very important situations where there are a few copies of specific chemical species" involved in particular biochemical interactions. As a result, "their interactions can't really be averaged to obtain an accurate and [in some cases] even a qualitatively accurate simulation."
For its part, "StochSS actually can solve ordinary differential equations but it goes a step further in providing discrete stochastic models that are targeted at basically smaller spatial scales," she said. Its approach is to consider "every reaction event along with its probability and then simulate each event virtually on the computer according to the probabilities." In many cases, "that leads to a much more accurate simulation that captures important properties of the system [which] a deterministic simulation like ODE or [partial differential equation] wouldn’t be able to do."
Petzold and her colleagues are using the NIH funds to develop what she described as an "integrated development environment," accessible by a user friendly web-interface, that provides researchers with tools to convert their cellular models — generated using ODEs, for example — into more complex discrete stochastic simulations. She and colleagues have been using their software to study cell polarization in yeast mating activity looking specifically at the biochemical interactions that occur when yeast cells of different mating types sense each other.
"We've got … powerful tools that the biologist could use to form what we call a spatial stochastic model where the actual spatial locations of the molecules … can be taken account of," she explained. "Normally, if you wanted to do that you'd have to learn a whole new package but here all you would have to learn is how to input the geometry."
Now in its third incarnation, StochSS currently has algorithms "for simulation and sensitivity analysis of ODE systems, simulation of well-mixed discrete stochastic systems, and parameter estimation for discrete stochastic systems," Petzold said, and by the fall of this year, the researchers hope to have added tools for "spatial stochastic simulation."
StochSS runs on standard laptops and desktops but users can move their computations to the cloud if they want to throw more resources at their simulations to cut down on computation time. Krintz's team is responsible for developing this part of the infrastructure.
"A key piece of our back end is to allow you to investigate and experiment with lots of different options and parameterizations of your model concurrently" and compare the results to identify the best model, she told BioInform. Using the cloud also allows scientists to share models within and across research groups. "StochSS facilitates that type of collaboration [and] makes it really simple," Krintz said. "So instead of everyone reinventing the wheel, we can leverage the work of others and make much faster progress."
Finally, the system is portable, Krintz said, meaning that scientists aren't locked in to a specific cloud vendor. Currently the system is compatible with Amazon web services, but it can be adapted to work with other systems developed by other companies as well as with private clouds as long as these platforms run the Linux operating system and meet a few other minor restrictions — for example they have to support virtualization, she said. For those looking to run the system locally, it works with Linux, Windows, and Apple laptops and desktops. Moving forward, the researchers hope to add support for other cloud systems besides Amazon.
"Our long term vision is to support lots of different types of simulations in the cloud. We are augmenting the system with quite a large diversity of simulation types but this can be extended to other simulation systems," Krintz said. "That's going to be really important going forward."