Last author

Susceptibility to many common conditions and diseases, from obesity to cancer, involves the combined effects of many genes and of the many environmental factors that influence how these genes function. On pages 423 and 429, Eric Schadt of Rosetta Inpharmatics in Seattle, Washington, and his colleagues detail an approach to identifying such 'gene networks' — an advance that could potentially lead to diagnostic markers and therapeutic targets for common diseases.

When did you realize a gene-network approach was needed?

During my graduate training in mathematics and molecular biology, I realized that many changes — and not just at the DNA level — lead to disease. When I started this work in 2000, very few people thought this way. Luckily, our collaborators at the Icelandic pharmaceutical company deCODE Genetics and at the University of California, Los Angeles, immediately joined our push towards a more holistic, gene-network approach.

Is the traditional, reductionist appoach of searching for 'the disease gene' pointless?

Absolutely not. Genome-wide association studies comparing individuals' genomes identify the dominant genetic variations that lead to disease — if not 'the disease gene'. We've developed statistical methods to determine connections among genes on the basis of their activities in different tissues. We then combine these findings from humans with mouse models of disease to assess how perturbations to gene-expression networks could cause disease traits.

What computational power is involved in this approach?

We needed high-performance computing equivalent to 7,000 central processing units. Without that scale of computational horsepower, we couldn't have done this work.

Is Merck, Rosetta's parent company, adopting this approach?

Yes. Once you know how a whole network of genes is perturbed, you can better assess the best points for therapeutic intervention.

Are clinicians collecting the samples necessary for these types of studies?

Not as quickly as we'd like, but once clinicians see the point they will do it. To uncover the relevant gene networks, we need complementary sets of disease and normal tissues. For example, one clinician is collecting matched liver, stomach and fat tissues from his gastric bypass surgeries. These samples will help us to determine why the severity of diabetes is often greatly reduced following surgery.