At A Glance:
Name: Kevin Gunderson
Position: Principal Scientist, Illumina since 1998
Previous Experience: Affymetrix 1996 1998
Education: PhD, biophysics, Stanford University, 1995
MS, physics, Stanford University
BS, physics, mathematics, St. Olaf's College
For most researchers, PCR-based assays like Applied Biosystems' TaqMan have been the primary route to SNP detection in the lab. Fed up with data-scalability issues and complex steps in existing array-based and PCR-based assays, a team led by Illumina's principal scientist Kevin Gunderson developed a whole-genome genotyping method that the company hopes will do to PCR assays what Lasik surgery did to contact lenses. It wasn't easy, though. The team spent three years tinkering with the method before publishing a paper in next month's issue of Nature Genetics entitled "Whole Genome Genotyping (WGG) on High Density DNA BeadArrays."
BioArray News spoke with Gunderson this week to discuss how the new method works, and what kind of impact it could have on the genomics space.
What is your position at Illumina?
I am a principal scientist responsible for leading a group to develop new assays, technologies for array-based genomics studies. The past few years I have been focusing on whole-genome applications and this Nature Genetics paper is one such development.
Was developing this kind of assay a high priority for Illumina?
Well, yeah, it was a vision that I had, and the company now shares that vision.
What inspired you to fulfill that vision?
I saw a pressing need for whole-genome genotyping assays to characterize variation across the genomes of many individuals within a population, and there was no assay to accomplish such a task. Genotyping has always been plagued by multiplexing issues, and I wanted a genotyping assay that mirrored an array-based gene expression experiment. Gene expression employs direct hybridization to arrays of the products generated from a single-tube sample preparation. You can read out tens of thousands of genes simultaneously. Ideally, a whole genome genotyping assay would be just as simple. We chose an array-based assay, obviously, since I work for an array company.
What are some of the minuses of using PCR amplification in SNP genotyping assays?
One issue [is] the scalability of the assay due to multiplexing issues. Every time you scale a PCR-based assay, you have to do some development to ensure there [are] no issues like degradation of data due to increased multiplexing; that's usually one of the primary concerns. It is just unclear whether or not PCR-based assays can scale to the whole genome. With the whole-genome assay, there is no scalability issue. Once you scale it for a few SNPs, it extends to an unlimited number of SNPs. It is really dependent in a way on the density of features on the array. Other concerns are carry-over contamination you have to be really meticulous to make sure you don't have carry-over. Finally, the overall yield of the PCR-based reaction is lower than our whole-genome amplification based assay.
Why do you think it has taken so long to produce an assay like this?
One of the largest challenges has been the complexity of the human genome. Unlike cDNA experiments, the complexity of the genome is over 50-fold greater because the cDNA complexity is usually 40 to 50 megabases, whereas we have 3,000 megabases for the genome. Higher complexity and the partial concentration of any loci is much lower in genomic samples. Only recently have methods for amplifying the whole genome been introduced. Finally, there was an ingrained belief that the human genome was to complex for array-based analysis. No one really knew if you could get single-base resolution by hybridizing genomic data to an array.
How long were you working on this?
We've been working on this for the past three years or so.
How is the whole-genome genotyping assay similar to a gene-expression experiment?
They are similar in the sense that they share a single-tube sample preparation, and hybridization to an array of capture probes. So you directly hybridize the amplified product and capture the loci, or genes, of interest. And in fact our genotyping arrays are very similar in design to our whole-genome expression arrays. Both use 50-mer probes to capture loci of interest.
What is the significance of using 50-mer oligos?
We chose 50-mers as a compromise between hybridization efficiency and length. We actually compared 35-mers, 50-mers, 70-mers, and going from 50 to 70 we didn't see a lot of difference in signal intensities.
You used the Sentrix bead array as your platform. Why did you choose them and how did they contribute to the quality of the assay?
Well I guess it is not really a choice because I work for a company that makes them.
You hypothetically could have brewed your own for the experiment…
That makes sense spotted arrays or something like that since we do make oligos. Bead arrays are of course an obvious choice, and they do have several advantages over conventional either spotted or in situ synthesized arrays. First of all, with bead arrays, there's only one historical immobilization event for each feature and we can make tens of thousands of arrays from that single immobilization event. There's no variance in array data due to features being compared from different immobilization events. Secondly we have higher redundancy -each data point consists of 30 beads. For each measurement we average about 30 beads. So that contributes to its robustness. In the case of in situ synthesized arrays it is really hard to generate full-length probes on the array surface.
So how did you validate your results using the WGG method versus data from real-time PCR-based assays?
Basically we took a set of SNPs, which were the HapMap QC SNPs, and we ran the Golden Gate assay on them and also ran the whole-genome genotyping assay and compared the quality of the data. The call rates, concordance, heritability, reproducibility, and so forth. In general the data was on par with Golden Gate quality.
Can you walk us through the method?
The WGG assay is composed of four modular steps. Each of these steps can be interchanged with alternatives. The first step is whole-genome amplification this makes a lot of DNA. This effectively increases the partial concentration of each locus and drives the hybridization capture. The second step is hybridization capture of the amplified genomic DNA, and it uses the 50-mer probes which bind adjacent to the SNP site. The third step is an array-based enzymatic allelic discrimination step. In particular, we employ allele-specific primer extension with labeled nucleotides. It's a well-known SNP-scoring strategy. The fourth step is signal amplification to increase the signal-to-noise ratio of the overall assay.
Have you submitted a patent application for this method?
We submitted a patent application almost two years ago.
Will you commercialize the assay?
Yes we will. The first product we are launching is the 100K Exon-centric Whole-Genome Genotyping Bead Chip and that should be released within the next few months. This product will be followed up with a pair of 250K HapMap "tagging SNP" Bead Chips. Finally, by June 2006 we will have two 500K Bead Chips for a total of over a million SNPs total.
Are your customers excited about the WGG method?
This assay is generating a lot of excitement. It ultimately reduces steps and costs. A number of researchers and organizations are gearing up to do whole-genome association studies that hadn't been possible until this technology came along. This is an exciting time in the whole-genome genotyping field also in pharmacogenomics. I see a future where all clinical trials will use a whole-genome genotyping product to characterize the variations in patients.