For a chromosome (chr 12) with only marginal genetic linkage signals, an unexpectedly large number of AD candidate genes have already been proposed over the past six years including A2M, LRP, OLR1, and LBP-1c. Now we need to add yet another, GAPD, which sits only 2.5 megabases from A2M. In this brute-force high-throughput genomics-style approach to genetics, over 200 SNPs were genotyped in a mixed set of case-control samples, and then tested for association with AD alone or taking into account gender and ApoE genotype status. In the end, this makes for a huge number of comparisons, thus raising the double-edged sword of testing hundreds of SNPs in a single linkage region; while this approach can be incredibly comprehensive, it also dramatically increases the probability of false positive results due to multiple comparisons, or as we sometimes call this in Boston, "The Curse of Bonferroni." Remember that 1 in 20 comparisons is likely to yield a false result. With over 200 SNPs genotyped and then tested for association with AD in multiple models, e.g., according to gender and/or...  Read more

For a chromosome (chr 12) with only marginal genetic linkage signals, an unexpectedly large number of AD candidate genes have already been proposed over the past six years including A2M, LRP, OLR1, and LBP-1c. Now we need to add yet another, GAPD, which sits only 2.5 megabases from A2M. In this brute-force high-throughput genomics-style approach to genetics, over 200 SNPs were genotyped in a mixed set of case-control samples, and then tested for association with AD alone or taking into account gender and ApoE genotype status. In the end, this makes for a huge number of comparisons, thus raising the double-edged sword of testing hundreds of SNPs in a single linkage region; while this approach can be incredibly comprehensive, it also dramatically increases the probability of false positive results due to multiple comparisons, or as we sometimes call this in Boston, "The Curse of Bonferroni." Remember that 1 in 20 comparisons is likely to yield a false result. With over 200 SNPs genotyped and then tested for association with AD in multiple models, e.g., according to gender and/or ApoE, a "meaningful" result would have to be, statistically speaking, very strong. The authors report odds ratios that are relatively small—all < 1.3. Thus, while there could be a meaningful result here, only further testing in more samples, case-control and family-based, will reveal whether GAPD will someday sit on a slide next to ApoE, or hover in limbo with many other controversial AD candidate genes. No one ever said solving the genetics of late-onset AD would be easy, even with the brute-force strength of high-throughput SNP genotyping.

View all comments by Rudy Tanzi

From the AlzGene team:
In addition to the three AD samples listed, this study also tested a fourth AD sample comprising subjects originating from a sib-pair linkage study. However, since these linkage AD cases were not contrasted to an independent sample of controls, we decided to not include it into the database.

View all comments by Lars Bertram