It has taken me some time to digest the impressions from the 1½ day International Imaging Genetics conference held in Irvine a couple of weeks ago. This is probably because it was hard to sort the different issues out initially. The conference had speakers from genetics, statistics and neuroimaging.
Correspondingly, there were three major Imaging Genetics (IG) themes one can sort this conference into: genetics methods, statistical approaches and visualisations, and neuroimaging related issues.
Doing statistics is a humbling experience, and the IG conference was a wonderful reminder of this. In the burgeoning field of IG, many studies that have been published in top rated journals would probably not even make it past the editor today. On the other hand, it’s only been a few years since one could get an article published in Nature or Science because you found some signal in the brain during a cognitive task. Anyway, the statistics were no exception to all the demonstrations associated with doing statistics, although it came with a twist. Since IG is a combination of at least two approaches – genetics and neuroimaging – each study must seek to accommodate to the pitfalls and premises of both approaches. This is not a simple task: neuroimaging contains a multitude of different statistical approaches, in addition to an overwhelming number of issues and pitfalls when it comes to the design, collection and preparation (i.e. preprocessing) stages in a study. I can only guess that the same goes for genetics.
During lunch, I heard a geneticist asking “What is a voxel? Is it like a pixel?”. So IG has a long way to go in order to reach a full, common understanding and sharing of ideas, concepts and methods. I’m certainly asking just the same kind of beginner’s questions about genetics. So when Bernie Devlin said about the speaker before him, Tom Nichols, “I am glad that Tom says he does not know much about genetics. I can assure you – he doesn’t!” he was making this very point.
But the IG statistics also had some very interesting and directly useful aspects. Nik Schork talked about different ways to visualise and analyse IG data, and demonstrated a most impressive toolbox of different methods for doing so. Unfortunately, I can’t find any illustrations online (nor in any article) to show this. I’ll get back with as soon as something comes up. See also this video of one of his talks.
This part was probably the hardest, since so much relied on one’s knowledge about genetics. Haplotype, SNPs, alleles and so on, just to mention a few. If you have not heard about this before, you’re not alone. But even knowing about these keywords and concepts, bringing them together with neuroimaging really poses a test of your working memory ability… I’ll expose my lack of understanding of these issues here,yet still mention the hapmap project and its tremendous usefulness in assessing the distribution of haplotypes in different populations. Not directly viable when doing neuroimaging studies, but it can influence the likelihood that you choose to study one haplotype rather than others.
This is by far the easiest part of the conference, at least to me, and I guess geneticists had the harder time in this part of the conference. However, one can also see that the neuroimaging studies that were presented here really demonstrated the end results of the tedious work that had been presented in the foregoing talks. Since neither the basics of neuroimaging signals, stats or pitfalls were presented as such, researchers from other – non-neuroimaging – approaches probably had an easier time than us genetics-nogoods had previously…
Basically, one could say that IG brings a new tool to look at what drives your neuroimaging data, even in healthy individuals. Studies by researchers as Ahmad Hariri, Dan Wainberger and Andreas Meyer-Lindenberg illustrates this point clearly. Their studies have now demonstrated that a natural variation in specific alleles produce different responses in not only the brains of the different subjects, but even how behaviour is affected. This includes a study of how long and short versions of a seretonin transporter gene affects brain regions affected in depression. It can also demonstrate how genes affect the brain to produce a higher risk of developing schizophrenia, or how a gene influences brain size. It can also be used to enhance our understanding of different cognitive functions, such as attentional networks, in the brain.
For last year’s conference you can now download video recordings and the slideshows of the talks from the Irvine IG conference homepage. I suspect that the talks for this year will be available soon, too.