I spent the past year sharing office space with Paul's lab, so this result doesn't really surprise me. I dislike the title that Technology Review is running, but that shouldn't detract from this.

The major histocompatibility complex sits on chromosome 6. Our best disease associations for HIV and autoimmune disease show up here. There are a number of published associations outside of this region, but nothing as blazing hot as the signal in the MHC. There's an ongoing debate in the field about 'missing variance' -- we can estimate the heritability of diseases haven't found enough variation in the genetic soup to explain the heritability.

(I'm afraid I need to run out the door, I'll do my best to answer questions on this work or similar studies.)

There are some groups within the Broad Institute that I think have the tools to more or less resolve the question of the missing heritability (genetic variance), and I'm sure you know these folks too. This work isn't published yet, so sadly we can't really talk about their answer.

I don't believe I'm privy to any secrets, but I'll do my best not to leak information.

My complaint is with additive models of liability. [In brief compute an independent relative risk for each SNP. To compute the risk of disease, take the log sum of the relative risks.] I haven't seen a model that can take into account epistatic effects.

But heritability expressly excludes non-additive factors so I don't think that should be a problem. I agree that non-additivity affects the overall variance.

How are they able to rule out variability in T cell recognition (responsible for the signaling that the cell should be destroyed) is allowing for immunity and not the MHC?

I haven't read through the whole paper yet (behind a paywall (I think... I'm at a university that subscribes, so I have access) here: http://www.sciencemag.org/cgi/content/abstract/science.11952...), but table 1 is somewhat telling: there is no single mutation that's found in all controllers, and not in progressors. Instead, there's a collection of SNPs that controllers have more often than progressors, but not all controllers have any given SNP.

You're almost never going to get perfect genetic discrimination in all-or-nothing fashion.

The final figure is very interesting: the key SNPs that affect controller status are located near one another in 3-space in the HLA-B binding groove at a key site for antigen presentation. The fact that the genetics are all pointing to specific parts of the molecule in 3-space is extremely helpful.

It would be interesting to know what would happen if one person had all the SNPs.

Would the person be able to cure HIV, or is there a reason there are only combinations present and not the full set (IE does the whole set destroy some other part of immune function)?