CONVERSATION | Affymetrix’s Stephen Fodor discusses building a tech business without a blueprint
INTERVIEW BY MALORYE A. BRANCA
July 14, 2004 | In 1989, physical chemist Stephen Fodor was recruited to join the Affymax Research Institute by renowned biochemist Lubert Stryer, and charged with the task of developing "high-density" microarrays that would eventually squeeze DNA oligonucleotides into microscopic grids. His breakthrough proof of principle took the cover of the Feb. 15, 1991, issue of Science.
Two years later, Affymetrix was born with Fodor as CEO. Fast-forward a decade: In 2003, Affymetrix achieved its first full year of profitability and was the first company to introduce an off-the-shelf, human whole-genome-expression array. Also last year, Affymetrix launched a pivotal new product line — the SNP analysis GeneChips — and began a diagnostic commercialization collaboration with Roche Diagnostics. Bio·IT World senior informatics editor Malorye A. Branca caught up with Fodor, inviting him to look back over the past decade — and ahead.
Q: Has it all turned out the way you thought it would?
A: Some of it has, but from the very beginning things changed dramatically and quickly. The first challenge was mainly a technical one: We had to fulfill the original vision of building a technology capable of analyzing the high-sequence content of the human genome. So we had to make an early, aggressive push into technology. We were trying to do things that had never been done before. Not only were we squeezing a lot more DNA onto a much tinier surface than ever before, we also had to build readers, the software, and everything else. Meanwhile, the commercial acceptance wasn't very high at the outset, so that had to be worked out.
What was the long-term vision?
That was the hardest part. I had many meetings with diagnostics companies that would say, 'There is no use for this technology. Why look at 100,000 things when I can get what I need from 10?' People had just identified things like [tumor suppressor gene] p53. And at that point, everyone was thinking we needed to understand only a few genes, and we'd need just a few markers for each one.
People were predicting the cancer diagnostics market alone would soon be worth $5 billion, but they couldn't see where the tools we were building fit in this picture. So there was a chasm between what we were doing and what the marketplace was telling us. We'd spent a lot of money already, and we weren't going to develop something with just 10 probes.
That was where we needed a lot of faith. I believed the real value would be in capturing the information content of the human genome, and I think that bet is paying off now. But at the time it was a real struggle, not just internally but with investors too. Some people wanted us to focus on very narrow products. So we split the company into diagnostics and genomics units.
The important thing is that the DNA chip isn't just another tool, like a reagent. It's an enabling technology. Like a microscope, it lets you see new things. Once you peek into the beyond, a lot of your ideas change.
So, what have you seen "in the beyond"?
For example, back in the early '90s I don't think the market appreciated that we'd be doing diagnosis and prognosis someday by looking at patterns of expressed genes. But based on the work in lymphoma and leukemia today, that's what a lot of people are interested in doing.
How did you manage to keep your heads during the genomics boom?
It's true there were a lot of temptations during the boom: You could create a company and raise $100 million just by sequencing your dog. As long as you put it on the Internet, everyone would get excited!
But the focus at Affymetrix has always been on real fundamentals. Thanks to the extraordinary scientists who helped start this — like Paul Berg and Lubert Stryer — we were focused on doing the best possible science.
Being too grounded in fundamentals may be one of my biggest weaknesses. But luckily, that's worked here. For example, when we bought Neomorphic in 2000, it had nothing to do with its bioinformatics business model, although bioinformatics was all the rage at the time. Rather, we wanted the company's expertise. These were the guys who helped with annotating the public version of the genome, and they've helped us do a better job of presenting the genome on our chips.
How did you handle the early problems with GeneChips having incorrect information?
In the late 1990s the databases were still fragmented, and that's all we had to work with. There was some mouse sequence information in the wrong orientation that ... was put on the chip. We'd always used the public sequence sources, but that experience made us realize we had to put much more rigorous quality control in place around when to accept a sequence. In fact, that was part of the rationale for the Neomorphic acquisition.
Besides putting more rigorous quality control in place, we took responsibility for the problem, and we were probably overgenerous in replacing the faulty chips.
How do you plan to stay ahead with so many competitors around?
One of our big advantages is that we already have the manufacturing issues worked out. There was no model for that. But our process has turned out to be very robust and easy to apply quality control measures to.
There is still a tremendous opportunity if we can truly present the entire genome on a chip, not just the genome as we now know it. The key to doing good science is to look beyond the places that people say you should look. And that is especially true for the genome. Clearly, it's not just in the genes we've identified where we will find interesting things. There is lots of information left to discover. Demystifying the genome is a huge opportunity for Affymetrix.
As more genomes come out, we'll have more specialized chips. Here, there are lots of competitors trying to come in on the low end and erode our business. One of the challenges is to organize so we can quickly launch new products. Genotyping is also a very important new area for us now.
At the same time, there is this whole clinical world, which brings us down completely new roads. Here, at least there was a model for us to follow — Intel's. There are already big companies that own most of the diagnostic market, so it doesn't make sense for us to go out and compete with the Abbotts and Roches. Instead, we're building the chip that plugs into other people's systems.
There has been tremendous progress in the last 10 years, but what hasn't changed is that you still have to do the experiment to learn something new.