Name: Malek Faham
Position: Director of Research, ParAllele, 2001 the present
Background: Research Associate at the Stanford Genome Technology Center and a Psychiatry Resident at the University of California San Francisco, 2000 2001
Education: PhD, University of California San Francisco, 1995 MD from the University of California San Francisco, 1997
The current issue of the Proceedings of the National Academy of Sciences has an article on mismatch repair detection, a technology that at least on the surface looks like it might prove to be a worthy resequencing contender. The method looks to be fast and accurate, or at least it seems that ParAllele BioScience feels that way 17 of the article's 20 authors hail from the South San Francisco, Calif.-based company.
Pharmacogenomics Reporter spoke to the article's lead author, Malek Faham, ParAllele's director of research, to learn more about the technology and how it compares to what scientists are already using.
ParAllele seems to be putting some effort into MRD. Is the company planning to do anything with the technology outside of research, perhaps diagnostics?
We are right now focused on the research use, and that could be research internally or, potentially, with customers using the technology. A high-throughput research tool.
What technology might it compete with?
There are other technologies out there that use resequencing, and they're kind of emerging, including single-molecule sequencing that multiple companies are trying to do, as well as the resequencing chip that Affymetrix has.
Can you compare this technology to the new sequencing technologies, such as sequencing by synthesis?
In a conceptual sense, they're both trying to achieve the same goal, which is basically to get sequence information, which gives you rare alleles in the human population that may contribute to disease or any other phenotype you're interested in for pharmacogenetics, or what have you.
They're all trying to do that, but the one that's going to win is going to have the highest performance metrics. It has to have a very low false positive, because the signal that is present the biological signal is by definition rare. And therefore, the false-positive rate has to be very low. So, we think that is going to be the case for our technology, and I can't yet say how other technology is going to be. We'll have to see.
Are there cost advantages over Sanger sequencing?
It depends on what scale we're talking about. For small scales, Sanger is better because the whole point of our technology is that you're doing hundreds of thousands of amplicons. So, we're doing that all in one reaction it's in one tube so the cost of our assay, whether you're doing a thousand or two, it's not that much different.
Obviously, for Sanger, the cost goes up linearly. In the lower scale, we're more expensive than Sanger. On the higher scale, we're significantly lower.
How much would it cost to analyze an entire mammalian genome?
One thing I want to point out is the difference between de novo sequencing and resequencing. Our technology is not a de novo sequencing technology; it's a resequencing technology, and it's a technology with a very low error rate.
The cost of one mammalian genome at what performance metric? That's always an important point, because you can get the 99 percent with some cost, but it's getting to 99.999 percent it's getting the last bit that's going to cost you $10,000 more than the first bit.
So, we put in oligo[nucleotides] to be able to resequence all of what we think of as the 'functional genome,' meaning the five to 10 percent of the genome that has some functional consequence. Our goal is to be able to do that for the cost of what people [spend] on whole-genome genotyping technologies now to get to the $1,000 genome, but it won't be the entire genome, it will be the five to ten percent that is relevant.
Was the number of amplicons 1,000 that you resequenced for the PNAS paper the important point was the most that had been done?
That has been the most. Well, we have done one exploratory [experiment] in yeast, but in humans, that has been the highest.
What were the important biological findings?
We confirmed previous experiments from various labs of the distribution of variance across exons in other words, we got a lot of rare variations and a small number of common variations. That observation is well known, and it's gratifying to have seen that using this technology.
The second thing is, I think we've validated a new approach to identify common disease. Instead of genotyping a set of samples for a known set of SNP that's one approach a second approach is sequencing a few individuals, and then genotype those SNPs that you find in those samples.
We took that to a new level, and said, 'Let's make the resequencing a screening technology.' That's the new approach using the resequencing as a screening approach to identify variants that are relevant in this case to autism. We identified several of them [SNPs and other mismatches], and obviously none of them is validated. They all need to be validated in independent populations, but that's something for the research community to follow up on.
But the main thing that we like to have shown is a new approach for identifying relevant variants in common disesase using resequencing as a screening tool for large scales for lots of patients and lots of controls.
Will this approach ever rival genotyping?
The main difference, and this really is the crux, is that in genotyping, you're assessing SNPs that are known. So, usually to be able to assess that, you're depending on common SNPs, and the common SNP-common disease hypothesis that common disease is caused by common SNPs. If that's the case, I'd say there's no advantage to a resequencing approach over a genotyping approach.
[That hypothesis] will be true in some cases there are no questions about that. What is also true and there are no questions about this either is that sometimes there are going to be rare or heterogeneous [variants]. And what we learn in genetics is, if it's possible, it will happen.
What a resequencing approach allows you, is to be able to be hypothesis-free about the allele frequency spectrum of the disease-causing variants are. It will identify if it's common, rare, or heterogeneous.
What disease areas is this resequencing best suited to studying? As a diagnostic?
We really don't know. I thought autism would be a good one, I still think it's a good one.
Because it's so complicated?
Because it's so complicated, because it's kind of a rare disease. It could be a common disease, though, like diabetes, or cardiovascular disease, or just being overweight. In all these [it has] been shown that rare variants do cause them.
I think one more interesting thing would be pharmacogenetics. If you think about disease genetics breast cancer, heart disease, whatever we know what the biological signal magnitude can be, because people have done linkage analysis for many years, and we know what kind of results they got. Which is basically that the signal is not that strong, so they can be caused by a common allele that is not very penetrant, or a rare allele that is more penetrant.
In pharmacogenetics, we don't know that. You can potentially have Mendelian situations you can have some variants that, if you have it, you have the side effect, if you don't have it, you don't have the side effect.
My point is that when you're dealing with pharmacogenetics, you have a small patient number, usually. Therefore, your only chance of finding it is if it's highly penetrant. If it's not, you have really very little chance to identify it. It is possible that we may be able to identify those rare, highly penetrant alleles that may relate to pharmacogenetic phenotypes in a large screening of many candidate genes.