The story has been corrected to reflect the proper name of a firm that three of the founders of Affomix had previously founded. The name of that company is Protometrix.
Antibody firm Affomix, founded about a year ago, is ready to launch a custom antibody service business using a technology that it said significantly improves the affinity and specificity of the antibodies while shortening the production time as compared to other methods such as yeast two-hybrid.
The company will begin providing the service commercially during the summer, after it has moved from its current 4,000-square-foot office to a 12,000-square-foot facility in New Haven, Conn., at the end of this month.
The company is also refining its technology so that it will be able to distinguish proteins that have undergone post-translational modifications from those that haven't.
The company, which launched in January 2008 with $7 million in venture capital, has performed work for one client in order to validate its approach, and is in discussions with other pharmaceutical and biotechnology firms and consortia, Mike Sherman, Affomix's president, told ProteoMonitor last week. Sherman is also a partner in Elm Street Ventures, one of the investors in Affomix.
Along with Sherman the other co-founders of Affomix are Michael Snyder and Sherman Weissman, both professors at Yale University and Michael Weiner, who has held various positions at a number of life science shops.
While antibody production is a relatively crowded business, it also is one with plenty of room for improvement. Indeed, two of the most cited reasons for the low adoption of antibody microarrays in proteomics research are the poor quality of antibodies and the low number of antibodies that are available.
Affomix, Sherman said, aims to address these bottlenecks by defining a so-called "affome" — an antibody equivalent to the proteome and genome.
"Our ultimate aim is to generate the affome, an antibody set against virtually all proteins in the human proteome," Sherman said. The plan is to do that in two steps: the first is to develop antibodies against at least two epitopes of every protein in the human proteome. Then, a second generation of its core technology will be used to select antibodies against polymorphisms, mutations, and post-translational modifications.
Ultimately, he added, the antibodies developed by Affomix, which are monoclonal, would allow researchers to measure and track proteins in a manner similar to NMR and mass spectrometry but in a cost-effective way.
Y2H That Discriminates
The basis for Affomix's technology is the yeast two-hybrid system. However, Weiner has also developed a proprietary step that improves antibody specificity. It is that method, Sherman said, that distinguishes Affomix from other antibody firms in the market.
The company starts with a library of antibodies generated in yeast cells, then uses Weiner's method to excise those antibodies that are non-specific. While Sherman spoke only in general terms about the technology, called Y2H Express, developed by Weiner, he said that it addresses one of the biggest issues when it comes to antibody development — antibodies that don't work.
The poor quality of antibodies is well known in the research field. In conference talks and interviews, Matthias Uhlén, a professor at the Royal Institute of Technology and the catalyst behind the Human Protein Atlas, has said that more than half of the commercially available antibodies the HPA has received for testing don't work.
The National Cancer Institute was concerned enough about the issue that in late 2007, it began an initiative, as part of its proteomics program, to improve antibody development and characterization by creating an antibody-characterization laboratory at its Frederick, Md., facility [See PM 11/29/07].
According to Sherman, about 95 percent of antibodies from a typical yeast two-hybrid study will be non-specific. Affomix has "gone through our library and rinsed out the antibodies that we know right from the get-go that are likely to be non-specific and that right away improves the chances that the antibodies we find will be specific to 30 percent," he said.
The company has done work with one client, which Sherman declined to identify, but that wanted to "kick the tires and … see how well we could perform." The client picked 10 proteins that it was interested in using, grouped into three categories: proteins for which antibodies that work already exist; those for which antibodies are commercially available but don't work properly; and those for which antibodies don't exist.
[ pagebreak ]
Affomix was able to develop antibodies against all 10 proteins, Sherman said, and is now ready to generate and send them to the client.
The other major problem with antibodies is that while there are plenty of them on the market, they target many of the same proteins. By some estimates, commercially available antibody arrays contain fewer than 600 antibodies. But Affomix's technology is not dependent on antigen quantity, which opens up its potential ability to generate antibodies against all proteins in the human proteome, according to Sherman.
"We are of the opinion that if you look at … enough epitopes of the protein, and if you do it properly," antibodies can be developed against all proteins, Sherman said. So far, he added, "we have not been stumped."
To date, Affomix has generated antibodies against about 250 protein targets.
Paul Lampe, a professor of molecular diagnostics at the Fred Hutchinson Cancer Research Center, said that developing antibodies against multiple epitopes, such as Affomix's approach, would reduce the chances for cross reactions.
"If you have two antibodies to the same thing and they both react to the given sample, then you would be more confident that it was real," he said, though he added that it would be a very expensive process.
Currently, Affomix has an automated assembly-line process capable of manufacturing antibodies against 150 targets per week, "and that's scaleable, [although] we don't need to scale it at the present time."
It takes three to four weeks from the time a customer puts in an order for Affomix to identify which part of the protein to target and to develop the antibody. By comparison, a comparable experiment using a classical yeast two-hybrid experiment would take three to four months, Sherman said.
While a customer has proprietary rights over how an antibody generated by Affomix is used, Sherman said that he expects Affomix either to have rights to the antibody itself, or that there would be no IP restrictions on them at all so that it can offer the antibodies to other customers.
Affomix is also looking to address a growing market in proteomics research, analysis of post-translational modifications. Mass spec-based methods such as electron-transfer dissociation and collision-induced dissociation have been developed for PTM work, but the Affomix technology would make the work cheaper and simpler, according to Sherman.
"If we're able to come up with a set of monoclonal antibodies … that allow customers to distinguish between modified and non-modified proteins and put that together with our protein quantitation technology, then … you wouldn't need mass-spec hardware or to send samples out for analysis by mass-spec companies," Sherman said. "The procedure would be straightforward and amenable to multiplexing.
"The way we're contemplating this somewhere down the pike, you'd apply your biological sample to an antibody capture array, or a pair of matched arrays, and in one fell swoop, measure how much of each protein you're looking for is present in the sample and then what portion of each of the proteins is post-translationally modified," he added.
Currently, while certain antibodies can disclose whether a protein has a phosphate group on its tyrosine, for example, no antibodies exist that can distinguish between modified and non-modified proteins.
"We want to provide a product and services that will allow people to do things in proteomics that they can't do … effectively with proteomics," Sherman said.
He declined to elaborate on the technology that Affomix has developed for PTM differentiation, but said it would be agnostic to the kind of modification. He added that he hopes it will be "ready for prime time" in a few months.
Lampe, however, was unsure how this could be readily addressed through the use of modification-specific antibodies.
"You need to have a specific antibody for the modification … either a phosphorylation site or an acetylation site," he said. "You can make ones that are generic for most proteins but a lot of these reagents [that have been able to identify a protein that has undergone post-translational modification] haven't been that great [with the exception of those to phosphotyrosine] because the conformation around those modifications affects the affinity."
If the founders of Affomix seem familiar, it's because they have extensive experience in the life sciences business: Sherman, Snyder, and Weissman are three-quarters of the team behind Protometrix, a company founded in 2001 to commercialize protein microarray technology developed by Snyder, one of the pioneers in the protein and antibody microarray field [See PM 01/28/02]. Proteometrix was purchased in 2004 by Invitrogen, now a division of Life Technologies, which sells the technology under the ProtoArray brand.
Meanwhile, Weiner has held various positions at Stratagene, 454 Life Sciences, and GlaxoSmithKline.
The company will eventually offer antibodies off-the-shelf, but it is in no rush to enter the catalog antibody business, Sherman said. "[It] is our ability to select antibodies in a very comprehensive and rapid and remarkably cost-effective manner that makes it practical to contemplate making a chip to measure 20,000 different proteins," he said. "We are in a service mode and we're telling customers, 'If you have an interest in generating antibodies against a lot of different targets and want to do it in a rapid and cost-effective manner, then come see us because we can help you.'"