Although investigators only have a limited understanding of the exact mechanisms by which RNA interference works, a group of biochemists and computer scientists at Virginia Polytechnic Institute have begun a project to use that knowledge to develop a computational model of the RNAi process in C. elegans.
According to Richard Helm, an associate professor of biochemistry at Virginia Tech, the research project is an outgrowth of his lab’s interest in quiescence, or the ability of cells to shut down metabolically and become dormant.
“We’ve been working with computer science people over a number of years trying to develop algorithms that could actually help predict what mechanisms could be used to explain these phenomena in organisms that are really good at shutting everything down yet” remain viable, Helm told RNAi News this week.
“Out of this came the idea that if we thought about this from an [RNA] interference perspective, we could use C. elegans as a model because there’s a lot of data out there,” he said. “We could house that data and begin to develop a model to predict what could be the outcome of a particular interference experiment.”
The starting point for such a model is a database of all relevant information, Alexey Onufriev, an assistant professor of computer science at Virginia Tech and a co-investigator on the research project, told RNAi News.
“We want to combine everything there is in the field,” he said. “There is a lot of data in the literature as to how [RNAi] works … [and] we want to combine it into one database.” This information, however, does have its limits, Onufriev noted.
“We don’t know the mechanism of the key steps [of RNAi],” he said. “For instance, we don’t exactly know how Dicer cuts dsRNA, [or] exactly how RISC works and how it cuts the mRNA.”
But some of the structures of the proteins involved in the RNAi process are available, and work on others is continuing, which is the key to the development of the computational model, Onufriev said.
“There are some pieces [of the RNAi machinery] which have been solved crystallographically,” Onufriev said. There are “lots of other pieces that have not been solved crystallographically, but there is some data as to roughly where they might sit [in relation to one another]. We will try to put this together, do some molecular modeling, and try to come up with” answers.
For instance, he noted that “the structures of the analogs of Dicer are known, and from the literature we can infer roughly how it is positioned. We will then use molecular dynamics to come up with plausible confirmations of the positions of [dsRNA] on Dicer and try to figure out which one is best.”
If the results of the model seem correct and jibe with other data out there, “we can have some reasonable atomic-resolution model of how this thing works, and can try to infer things about the mechanism,” Onufriev said. “The same thing with RISC; the key protein of RISC, argonaute, has been solved crystallographically, and that’s going to be the starting point for us” there, he added.
“The grand, ambitious goal is to make a computational model of how [RNAi] works,” Onufriev said. “Suppose you have a virus, and you ask yourself, ‘What RNAi sequence do I want … that will have the maximum likelihood of silencing that virus or that particular gene?’ [If] you have this computational model, you feed in what you want to target, and you get sequences, a bunch of plausible answers that are your best bets,” he said.
In terms of answers about the structure of different RNAi mechanisms, Onufriev noted that “anything we can get out of [the model] will be good because there is really nothing available at this stage. In that sense, it’s difficult because there is little available; on the other hand it’s good because we will be one of the first ones in the field, at least at the molecular modeling level,” he added.
Despite the ongoing efforts of other scientists to determine the structure of the RNAi machinery, Onufriev still sees a place for the computational model.
“It may be that within a few years people will come up with the structures, or maybe not,” Onufriev said. “It’s really hard to say — this is an art, and sometimes things are not crystallizable or hard to crystallize, and people cannot do it.”
Additionally, the crystallization of a complex can be quite challenging, he said. “Let’s say you have the whole RISC [crystallized], but most likely it will be without RNA,” Onufriev explained. “So where does RNA sit — how do you position it? That question will likely remain even if” RISC itself has been crystallized.
Once a working version of the computational model is developed, it will be handed off to Helm and his colleagues for validation.
Using the model, “we can … say, ‘If we knock down expression of gene Z, we should expect protein levels to increase over here, and transcription to be modified over here,’” Helm said. “Then we can go ahead and test that biologically at the bench and see how the model comes out.”
According to Helm, the researchers hope that the model can be used as a “hypothesis generator” of sorts.
“I can’t look at all the gene-expression data [from] RNAi experiments that have been done in C. elegans and come up with a global picture,” he said. “That’s where the computer comes in — to help me model what this organism is doing. From that, [we can hopefully] get a hypothesis that we can actually test.”
Helm said that he and his colleagues plan to test hypotheses related to their expertise in longevity and aging. “We would like to focus on how an organism responds [to various stresses] to enhance its lifespan,” he said.
“This gets into the larger concept of a systems biology approach to an organism — how does it respond to a multi-stress environment, and if I modify key genes as part of that, can I predict its response,” Helm added.
The development of the computer model is being supported by a four-year National Science Foundation grant worth almost $1 million, Helm said. In addition to Helm and Onufriev, co-investigators on the project include Lenwood Heath, professor of computer science at Virginia Tech; Malcolm Potts, professor of biochemistry at Virginia Tech; Naren Ramakrishnan, associate professor of computer science at Virginia Tech; Salvatore Paxia, a researcher at New York University’s Courant Institute for Mathematical Sciences; and Edmond Schonberg, a professor of computer science at NYU.
As for when it may be ready for testing, Onufriev said that “the NSF has funded us for four years, so we have to have something by then [and] I think that’s a realistic goal.
“I would say by the mid-third year we should have something … some prototype that will do something,” he added. “It will be something to start with … and hopefully we’ll get more funding or other people will take it from there.”