It's fitting that EpiVax occupies a niche area in the bioinformatics market, because finding molecular niches is what its tools do best.
The company, founded in 1998 by Anne De Groot, director of the Brown University TB/HIV research lab, is one of only six or seven organizations worldwide that specialize in what De Groot calls “computational immunology.” Specifically, EpiVax has developed bioinformatics software to sniff out epitopes — localized regions on the surface of antigens that elicit an immune response and are useful for creating vaccines, diagnostics, and therapeutics.
EpiVax, which operates on a services-based model, has recently tested the effectiveness of its approach using the genome sequence data for the SARS virus. “We took the genome that was published and three days later had identified epitopes that we are going to put in a vaccine,” De Groot said. The company has filed provisional patents on the predicted epitopes.
De Groot said that the company verifies all of its computational predictions via in vitro screening in human T-cells and transgenic mice, and has an average accuracy rate of 85 percent. In addition to SARS, EpiVax also has demonstration projects in tuberculosis, HIV, human papilloma virus, West Nile Virus, and Epstein-Barr virus. EpiVax is currently in the process of experimentally validating the predicted SARS epitopes.
The company’s computational platform, called EpiMatrix, uses a combination of approaches to predict epitopes. The primary method is based on statistically determined sequence motifs in groups of peptides that are known to bind to a particular human leukocyte antigen. “When we build the programs, we’re looking at amino acid motifs that are recurrent in populations of peptides that are known to bind to a given HLA,” De Groot said. Another approach, which EpiVax is turning to now, begins with the structure of the HLA molecule and then predicts the peptides that will bind to it. This approach, called the pocket profile method, has been found to work across species, De Groot said, so EpiVax is using it to build prediction tools for swine and cattle as well as the human immune system.
De Groot said the company has built epitope prediction models for the thousands of classes of HLA molecules in the human population. The “secret ingredient” to the company’s approach is the quality of the data used to build the models, she said — “the information that you put into the program that detects those patterns for each of those hundreds of different HLA molecules.” The company’s hybrid informatics and wetlab process creates a “feedback loop” that helps refine the model, she said. Predictions that are proven to be accurate at the bench are confirmed, while those that are found to be inaccurate are improved.
In the case of HLA-A2, for example, which is found in about 20-30 percent of the population, De Groot said the company can run a genome through that particular model to determine what the T-cells of that population would “see” — which particular peptides would be presented to their immune system. This information could be used as the basis for vaccines or diagnostics, she said.
The approach also has applications in the realm of therapeutics, De Groot said: “A lot of the genes that people clone, like erythropoietin, thrombopoietin, and growth hormones, when they put them back into people, they generate an immune response, so [some customers] want us to map the epitopes in those compounds and take them out.”
The company is constantly updating its platform to match new data, De Groot said. For example, the company found that in the case of HIV, although the virus only has nine open reading frames, “there are literally hundreds of thousands of variants of HIV, so then the problem becomes finding not only the region that is going to turn on a T-cell, but the region that is conserved in all these different strains.” In response, the company built another pattern-matching tool called Conservatrix to find conserved regions across strains — a tool that came in handy for the highly variable SARS genome as well, De Groot said.
EpiVax has calculated that its computational approach offers a 6-fold to 20-fold savings in terms of cost and manpower over manual methods of mapping epitopes, which involve overlapping synthetic 15-mers across the length of an entire protein.
De Groot said that the six-person company is currently working with “five of the larger companies in the US that make either vaccines or therapeutic products.”