PHOENIX, October 27, 2005 — Researchers studying the genetic roots of autism at the Translational Genomics Research Institute (TGen) in Phoenix, Ariz., recently completed genotyping of 7,500 samples from 1,500 families using the Affymetrix 10K microarray. Their work is part of the National Alliance for Autism Research (NAAR) Autism Genome Project, a worldwide collaboration involving 43 institutions in the largest genetic study of autism ever conducted. In a parallel, but non-overlapping study, TGen is also performing whole-genome association studies on the publicly available National Institute of Mental Health autism sample collection with the 500K microarray.

A multitude of other complex genetic disorders are also being deciphered at TGen using the 500K array. They include Alzheimer’s disease, multiple sclerosis, progressive supranuclear palsy, memory defects, and bipolar disease.

Dietrich Stephan, director of the Neurogenomics Division at TGen, spoke with AMB Editor Rachel Shreter about combining newly-published HapMap data with 500K genotype data to complete the analysis of their whole-genome association studies and pinpoint the disease genes.

Shreter: Could you describe the autism association study you are conducting?

Stephan: We’re pinpointing new regions of the genome which harbor autism-susceptibility genes. Using a multi-pronged approach of performing genotyping for the Autism Genome Project consortium in a huge set of multiplex pedigrees, as well as a high-density whole-genome association study in our own lab, we, as a community, hope to arrive at a set of allelic variants of genes that cause a heritable predisposition to the trait. We know there are probably multiple genetic lesions that either alone, or in combination, predispose someone to autism.

In order to get around the issue of genetic heterogeneity, the Autism Genome Project consortium is performing a huge linkage study, powered such that subgroups that are caused by specific individual genetic lesions fall out. This is a collaboration that was spearheaded by a group called NAAR, but really encompassed the entire world’s collection of autism families—1,500 families—all of which had more than two individuals affected. In my own lab we were able to obtain the publically available NIMH autism sample collection because we are AGRE and NIMH approved investigators, and we took one proband from each of those families and did whole-genome association using the 500K chip.

We plan to deposit our 500k data into the public domain immediately upon publication so that data from the community 10k linkage study can be enhanced by our independent efforts. Although my lab is performing the genotyping, any and all 10k analyses are being performed by the consortium and not at TGen.

Whole-genome association studies using the 500K platform—or any high-density SNP genotyping technology—are going to require a lot of samples, probably more samples than any one person can collect. We’ve had to turn, in every case, to large national or international clinical consortia. For the autism study, the Autism Genome Project is using samples from the ten repositories. For Alzheimer’s disease, my team drew on the national network of 30 Alzheimer’s disease centers, each of which has a brain bank that’s been collecting samples for the past 15 years. In addition, we obtained samples from the Netherlands, a German brain bank, and a

Swiss sample cohort. Collectively, we have 1,000 cases and 1,000 controls meeting the clinical and histopathologic criteria that we set forth. Granular and accurate phenotyping are crucial. So that’s the same scenario that we’ve seen over and over again across disease areas—scientists banding together to find alleles of genes for common human diseases. The key to these things, at least on the screening side, is to get enough samples so that the study is powered correctly and you can get a real answer out of your training set.

We’re also using pooling-based approaches that yield allelic imbalances of more than five percent within months, as opposed to years. Of course, it becomes impossible to shuffle the comparison cohorts to find, for example, associations with multiple clinical parameters. So rather than do linkage in pedigrees and watch how things are inherited, we’re just going to see if there is an allelic imbalance between hundreds of cases, versus hundreds of controls at any locus.

Shreter: What about the analysis of these data?

Stephan: The analysis portion of the project is something that the whole field is wading through right now. That’s where the HapMap comes into play. We’re in the middle of trying to overlay all of the haplotype blocks out of the HapMap project over the 500K SNP map to see which SNPs define which blocks, to determine the