9 September 2011. Astute readers may have noticed that the 40 or so genes once populating AlzGene’s Top Results list have been culled to 10, and that former hits such as SORL1 and GAB2 no longer make the current cut. What happened, and why?
“We have changed the threshold criteria for what we declare to be a 'top result,'" said Lars Bertram of the Max-Planck Institute for Molecular Genetics in Berlin, Germany, who heads AlzGene’s team of curators. When the database was launched (Bertram et al., 2007), a gene variant made the list if it met the minimum standard for statistical significance, i.e., p-value <0.05, in an Alzheimer’s disease genetic association study. However, “with the advent of genomewide association studies (GWAS) that test millions of genetic markers per sample, that threshold became useless,” Bertram told ARF. Polymorphisms with genomewide significance typically have p-values ~5 E-08 (0.00000005)—six orders of magnitude from the nominal standard.
The AlzGene team adjusted the Top Results inclusion criteria to reflect the higher stringency this spring after the most recent AD GWAS were published (ARF related news story on Naj et al., 2011 and Hollingworth et al., 2011). As explained in Top Results Details (i.e., the link found just above the list of genes), Bertram and colleagues set the new threshold as E-05 (0.00001). That is an "in-between" solution—between the nominal and GWAS standards, Bertram said, adding that “the point is to avoid false-positive top results, but not be too stringent to exclude everything.” Genes falling short of the new standard are still in the database, accessible by searching under "Gene," "Protein," or other headings on the AlzGene homepage, or by e-mailing the curators. “They are just no longer highlighted as super interesting,” Bertram said. He also noted that the Top Results list is not a "Top 10" list. “If there were 100 meeting the new criterion, we would list them all,” he said.
AD geneticists find the change helpful. “This presents to the AD community, particularly to non-geneticists, a list of genes for which we have high confidence that they are really involved in AD risk. That is useful, because we couldn’t say that about the old list,” said Alison Goate of Washington University School of Medicine in St. Louis, Missouri. Though it is still possible that real genes may lurk among those dropped from the list because they did not meet the new cutoff, she said, “It’s better to be conservative and say ‘these are the ones we truly believe are involved in AD,’ so that if others want to work on the biology, these are the genes to focus on.” John Hardy of University College London, U.K., thinks the bar could have been set even higher, to E-07 (0.0000001), for the Top Results list. “This is genomewide significant,” Hardy wrote in an e-mail to ARF. “I think all better than E-07 will turn out to be right.” All current AlzGene Top Results exceed this more stringent threshold by at least two orders of magnitude, meaning that all show p-values of ~3 E-09 (0.000000003).
To be sure, GWAS are not the be-all-end-all of AD genetics research. “One thing people in the field wonder is, Well, how far do we take this? Do we need 100,000 samples? Do we need a million?” Goate said. The latest GWAS involved 7,000 to 8,000 AD cases each, and an international collaboration was launched earlier this year to pool data from French, British, and U.S. cohorts into a mega-GWAS with some 20,000 AD patients and 20,000 controls (ARF related news story). Nevertheless, Hardy said “We’re getting toward the end of the period where GWAS are big news.” In Goate’s view, “There probably isn’t a huge need to go beyond 30,000 samples.”
She did note one sticking point—published AD GWAS have thus far been limited to people of European origin. “We badly need GWAS in other populations,” Goate said, noting that research in other diseases has uncovered population-dependent gene effects. The Alzheimer’s Disease Genetics Consortium (ADGC) has an ongoing study involving several thousand African-Americans with results expected near the end of the year, Goate said.
While genomewide analyses approach what some see as their last hurrah in the AD field, next-generation sequencing methods are coming to the fore (see ARF related news story). Sequencing costs dropped more than 40,000-fold between 2005 and 2010. If the trend continues, exome and whole-genome sequencing may become cost-effective approaches for screening thousands of sporadic AD cases. Even with their current cost of $2,000 to $4,000 a pop, Goate said, these methods could be useful for finding additional rare mutations that cause familial AD. Occurring in less than 1 percent of the population, such variants—like those already identified in amyloid precursor protein (APP) and presenilin (PS)—are too rare to turn up in a GWAS, but whole-genome or whole-exome sequencing could “find them easily,” Goate said. Several months ago, scientists using exome sequencing reported new mutations in a cargo sorting protein (VPS35) involved in Parkinson’s disease (ARF related news story on Zimprich et al., 2011 and Vilariño-Güell et al., 2011). “While rare mutations are not the focus of AlzGene, they are collected in a separate database curated by Marc Cruts at the University of Antwerp in Belgium,” Bertram said. “We are getting ready to systematically include next-generation sequencing data on AlzGene if the frequency of the identified variants exceeds at least 1 percent in the general population, which makes them informative in the setting of association studies.”
Ultimately, GWAS and next-generation sequencing are merely “the first step,” Bertram said. “You have to drill down and find out what it is about this gene that makes it a risk factor for AD. What does the protein do in the brain that changes during disease? What particular change in the DNA sequence is responsible for that? Does it have any bearing on diagnosis? These methods are a powerful leap, but by no means the last step.” For example, a paper published August 31 in Nature online offers a way to connect biological mechanisms with GWAS data in a different field. An international team of scientists identified 37 new genetic variants associated with blood concentrations of metabolites involved in various common disorders including cardiovascular and kidney disease, type 2 diabetes, and cancer (Suhre et al., 2011).—Esther Landhuis.