As array density continues to increase and efforts like the 1000 Genomes Project churn out new variation data, the array community has begun to debate the optimal chip design for the next round of genome-wide association studies. Will chips with broad coverage of multiple populations be favored, as in the past, or will chips containing variation specific to certain populations be considered a better approach?
With the expectation that some of the next round of genome-wide association studies will be driven by new population-specific arrays, vendors are positioning themselves to offer customers conducting such studies both custom and catalog options. Still, some believe that multi-population "cosmopolitan" chips will be more than sufficient for most GWAS going forward.
Affymetrix has already made genome-wide arrays available for studying European and East Asian populations on its next-generation Axiom genotyping platform. A Chinese population-specific custom product, called the Axiom Chinese myDesign Genotyping Array, will be made available next year, the company said (BAN 8/3/2010).
Jay Kaufman, vice president of DNA product marketing at Affymetrix, said last week that the Santa Clara, Calif.-based firm will soon launch other population-specific arrays based on content generated by a large genotyping project led by researchers at University of California, San Francisco, and Kaiser Permanente.
Kaiser and UCSF researchers last year said the project, which will generate four ethnic-specific arrays, aims to genotype 100,000 Kaiser volunteers by the middle of next year (BAN 10/29/2010).
In addition to the European and Asian designs, Affy will commercialize arrays for studying African-American and Latino populations. "Those are all going to be used to genotype different numbers of samples for the Kaiser-UCSF project, but Affymetrix has the right to sell those commercially to other researchers in 2011," Kaufman said.
Kaufman spoke to BioArray News at the American Society of Human Genetics annual meeting in Washington, DC, last week. During the conference, Pui-Yan Kwok, a professor at UCSF involved in the Kaiser study, stressed the need to run population-focused chips in the next round of GWAS, as opposed to the more generic chips that have been used to date.
"For association studies, you are trying to match cases and controls so that the only difference between them is their disease," Kwok told BioArray News. "But if you don't match them with the ethnic background properly, then you will be getting hits the wrong way and you will be overwhelmed by the structural differences between the two populations."
Kwok predicted that, given the output of 1000 Genomes as well as the Kaiser-UCSF study, most researchers conducting GWAS will run population-specific studies going forward. "I think that people that know what they are doing will do it this way, though a lot of groups out there just pick whatever's on the shelf," he said. "Once the information is available, more people will do more selective studies."
'The Population Bottleneck'
The increasing capability to run population-specific GWA studies is being supported in part by two factors, according to several sources interviewed at ASHG. One is the availability of higher-density chips, and both Affy and Illumina have pledged to eventually deliver 5-million-marker chips to the market. The other is the wealth of content being generated by 1000 Genomes; the UCSF-Kaiser alliance; internally at Affy; and from a number of other sources.
The arrays used in GWA studies to date were largely based on data released by the International Human HapMap Project. That project genotyped 270 samples from a handful of populations, including parent-child trios from the Yoruba people of Nigeria and Americans of European ancestry, as well unrelated individuals from Japan and China.
In contrast, the 1000 Genomes Project, which recently finished its pilot phase, expects by the end of next year to complete sequencing 2,500 individuals of nearly 30 different population groups. While first-generation, whole-genome arrays were able to cover most of the content generated by the HapMap Project, higher-density products are now required to successfully cover all known human genetic variation with one product.
For instance, while Illumina's OmniExpress BeadChip covered 95 percent of known common variation following the release of the HapMap data, that chip only represents about 60 percent of known content today. According to Illumina bioinformatics scientist Michael Eberle, who discussed the issue during the conference, the company's Omni2.5 BeadChips, which contains 2.5 million markers per sample and were launched in June, cover around 86 percent of the content currently available from the 1000 Genomes Project.
[ pagebreak ]
Still, while enough data now exists to run association studies on populations of interest using high-density chips, it is unclear what approach will best suit the market.
According to Jeff Barrett, an expert on disease genetics at the Wellcome Trust Sanger Institute, the choice between a more generic array and a population-specific array may ultimately be determined by the population being studied.
"It depends on where your samples come from," he told BioArray News. "For Europe and Asia, and to some extent Southeast Asia, the [Illumina] Omni2.5 and 5 BeadChips are going to be pretty much all you want for any of those populations," he said. At the same time, he said that researchers studying African populations will require population-specific arrays for their studies.
"In Africa the challenge is going to be greater," Barrett said. "I don't think you can with even 5 million SNPs design a good African chip because the amount of variation within Africa is much greater than all the variation within the entire rest of the world," he said, noting that studies of African Americans will likely pose similar challenges.
"The population bottleneck out of Africa deleted the variation that still exists within Africa from the rest of the world," Barrett said. "Some of these groups have been living geographically close to each other for thousands of years but are still very genetically distinct."
Barrett estimated that there are about 10 ethnic groups that he and fellow researchers wish to study in Africa. Kirk Rockett, a researcher at the Wellcome Trust Centre for Human Genetics at the University of Oxford in the UK, is working with Barrett on MalariaGen, a multinational research project studying the genetics of malaria in Africa.
Rockett also said that the 1000 Genomes Project will give researchers studying African populations the data to improve their studies. "Once more and more populations are sequenced, we will see that the SNPs within populations are different, and that will form the basis for custom chips," Rockett told BioArray News.
"The problem with all these populations is that you find a SNP in one population and then you try to transfer that SNP to other populations and you don't find it," he added. "And is that because it's not a real association, or because the SNP you are typing is in linkage with function in African populations but not in others?"
Illumina has sold a specific array, the Sentrix HumanHap650Y Genotyping BeadChip, for GWA studies based on the Yoruban content from the HapMap project. Rockett said that the 650Y was "better" than previous chips but that the release of the Omni2.5, which includes 1000 Genomes data, has gotten the project "closer to the density we want."
Going forward, Rockett speculated that Illumina could make versions of the Omni2.5 that are specific to populations of interest. "I think it will be like the Omni2.5, but a different stripe that is specific to a certain population," Rockett said.
Cosmopolitan vs. Custom
According to Omead Ostadan, vice president of marketing at Illumina, the company is currently assessing whether a "cosmopolitan" chip, like the Omni2.5 or Omni5, will be sufficient for most researchers conducting second-wave GWA studies, or if they will require more population-specific chips. Ostadan said the San Diego company's scientists would likely be able to make a decision on the issue once more information from the 1000 Genomes Project becomes available.
"If there is a need, that is something we should be able to determine in silico," Ostadan told BioArray News. "Illumina has the fundamental capability to create population-specific chips."
Carsten Rosenow, Illumina's associate director of global market development for genotyping, held a similar view. "Illumina can do any population-specific array if there's a need for it," he told BioArray News at ASHG. "The current thinking is that for all SNPs with a minor allele frequency under 2.5 perfect, it's still cross-population comparable," Rosenow said. "The assumption right now … is that we will see many more monomorphic, private SNPs, so the thinking is that for the next round you take a population-specific approach.
"It really depends on what comes out of 1000 Genomes," he added. "We have to look at the data and see what the SNPs are and then make a decision on what the best way forward is."
Affymetrix, in addition to the population-specific catalog arrays it has launched and intends to launch, will similarly allow customers to design arrays using its MyDesign program. Affy's Kaufman said that customers already can use available data via a web interface and submit a SNP list to the firm to order custom chips. "We can also engage with you if you want to do something a little more off the beaten path," he noted.
The University of Oxford's Rockett said that, for now, both of the major vendors should be able to serve the GWAS market by selling both catalog and custom population-specific arrays. "I think the vendors are doing everything they can do," Rockett said. "They are trying to design chips everyone can use, and will design custom chips."
According to Rockett, it "might be that you start off with a broader array, and then through sequencing or using 1000 Genomes data on that population, you can go and design your own chip." One of the things he and fellow researchers are hoping, though, is that "maybe in five years' time, when the technology comes along, we'll just do our GWAS with sequencing."
Have topics you'd like to see covered in BioArray News? Contact the editor at jpetrone [at] genomeweb [dot] com