Natural Selection, a company that uses so-called computational intelligence methods to develop pattern recognition tools solutions in areas ranging from factory scheduling to breast cancer detection, expects to begin licensing it internally developed microRNA gene-detection algorithms as early as the first quarter 2006, RNAi News has learned.
When it was founded in 1993, Natural Selection's primary focus was defense contracting, Gary Fogel, vice president of the La Jolla, Calif.-based company, told RNAi News last week. However, after Fogel joined Natural Selection (which was co-founded by his father) after receiving his PhD in molecular biology from the University of California, Los Angeles, the company began looking into applying its technologies to the life sciences.
"Seeing RNAi and microRNAs take off in the last few years, I figured [that it is] an area we should get into," he said.
In line with that goal, the company applied for and received in July last year a National Science Foundation phase I SBIR grant, worth almost $100,000, to develop a method for detecting microRNAs using neural networks, which are essentially pattern-recognition models designed to enable researchers to evaluate different statistics together or in different combinations (see RNAi News, 6/18/2004).
"The idea was to find a pattern-recognition model that would discriminate between genes that that code for functional RNAs, such as tRNAs [or] rRNAs … and non-functional RNA genes," Fogel said. "We did that with very high success on C. elegans, human, mouse, and rat."
"Seeing RNAi and microRNAs take off in the last few years, I figured [that it is] an area we should get into."
According to Fogel, the next step was to further develop the algorithm for specific RNA types, namely miRNAs, in a specific organism.
Earlier this year, the company was awarded a phase II SBIR grant from the NSF, worth about $400,000, to develop software to identify functional RNAs in the human, mouse, rat, C. elegans, Drosophila, zebrafish, chicken, Arabidopsis, corn, and rice. "The commercial application of this project will be to identify a new class of targets for drug design and discovery for the pharmaceutical industry," the grant abstract states.
Fogel noted that while the company's technology can be applied to all functional RNAs, Natural Selection is focusing on miRNAs "because that's our audience of interest. If people are interested in finding tRNA genes, although we don't think they really are, we could do that."
Fogel said that he has pinpointed the first quarter next year as the possible launch time for the miRNA detection software based on "the rate we're going."
He noted that the company is planning on first beta-testing the software in collaboration with industry and academic partners in order to validate its effectiveness, "and we have to do that validation experimentally, which is something that is part of our phase II [grant project]. Once we've demonstrated the efficacy of the [software's] discrimination [ability], we're going to start licensing the models."
Fogel said that validation would likely be limited to the human and mouse versions of the miRNA detection software given the expense involved, "but I think that if we have a good handle on it working in human and mouse, we'll certainly have it for mammals. Getting into plants might be a different story," he added, "and I'm talking with people right now about how we can get some experimental assistance in the plant area."
As for what happens after it's launched its miRNA detection software, Fogel said that that Natural Selection has its eye on tools for pinpointing miRNA targets.
"The idea was to find a pattern-recognition model that would discriminate between genes that that code for functional RNAs, such as tRNAs [or] rRNAs … and non-functional RNA genes. We did that with very high success on C. elegans, human, mouse, and rat."
"One thing is to find where these microRNAs are being made, another is to find what their targets are," he said. "We're interested in this as well, and extending this further to not just gene identification but target identification would be a future direction we're going to be headed."
However, Fogel does not expect this to be an easy endeavor.
"Given a database of known, experimentally verified microRNA genes, its relatively easy … to develop a model that can understand those data and make predictions about future microRNA genes," he said.
"When you then go to the output from that gene and where it might bind … it's clear that there are lots of different places where small RNAs can bind and bind with some success," Fogel said. "There are so many more possible solutions in the space to search through, to find the right one is a much bigger challenge, especially when we don't have the database yet of lots of different targets and their bindings.
"Before the experimental evidence piles up, which would be wonderful to have, it's a difficult problem simply because the search base explodes on you," Fogel added. "Once the experimental evidence becomes easier to generate, then it will be easier for us to generate our model."
— Doug Macron ([email protected])