The American Society of Human Genetics (ASHG, pronounced “A-shag”), annual meeting is a scientific, networking and socializing milestone every year. This year the meeting was 8 time zones away from the UK in San Francisco, CA. It was a busy year for our group, with 4 talks and a poster being presented, a variety of collaborators’ meetings to attend, a choice of hundreds of talks to listen to, and, of course, plenty of drinking to do.

Each member of the group who was there offers their thoughts after the break. It’s interesting to see that while we covered a wide variety of topics across the group, the most consistent message is that this year didn’t yield any major discoveries that will change the field. Instead, we all saw incremental progress in applying next-gen sequencing and similar technologies to many different problems. Perhaps this is simply a reflection of the nature of modern human genetics: a gradual improvement of our understanding, rather than a sudden revelation.

One of my favorite talks this year was the one from the GTEx Consortium (presented by Kristin Ardlie). Their main aim is to identify the genomic regions that influence whether and how much a gene is expressed. They do this by correlating genotype data and tissue-specific gene expression levels. The tissues are mostly collected from deceased individuals. At first this sounded a bit grim, but it gave them the opportunity to collect a large set of samples, and otherwise difficult to obtain tissues (brain, pancreas, …). They also assessed the effects of different ischemic times on expression levels. They are still collecting data, and all (including methodologies) will be made publicly available.

In general, while last year’s ASHG was clearly focused around the if and how of whole-genome sequencing, I did not experience a common theme to this year’s meeting. It was clear though that many groups had embarked on whole-genome or whole-exome sequencing in the previous year, and were ready to present their results. Several sessions therefore felt as a list of studies where the same technique was applied to yet another disease. Although each of these undoubtedly are of extreme importance, what I sometimes missed was the ‘what after locus discovery’ part. But it is still the early days of the whole next-generation sequencing adventure. I feel we are now somewhere on the tipping point between merely employing these techniques on the one hand, and translating their outcomes into insights in disease pathogenesis on the other (for example by integrating with eQTL analyses and performing functional studies). I was therefore very pleased to see that the organizers had included a session on the functional follow-up of associated variants in complex disease.

As developing bioinformatics tools is my day job I am always interested to see what all the other bioinformaticians working on human genetics are up to. From the bioinformatics posters and commercial booths it was clear that pipelines and tools for annotation, filtering and prioritisation of variants from next generation sequencing techniques is a common problem that many people are working on. This is reassuring as version 1.0 of our own variant filtering tool for family sequencing studies Olorin was recently released.

Within these tools there seemed a clear focus towards performing these tasks in clinical settings, where bioinformatics expertise and computing resources may be limited. A common solution to the problem of limited compute resources was the creation of cloud-based workflows. This is an area I would like to get more involved in as it seems like a perfect way to make the analysis pipelines that we are all creating available for the wider scientific community. It will be interesting to see which if any of these tools get the most traction and become the go to tool for filtering sequencing variants.

This was my first time giving a presentation at the ASHG. Although it was very stressful getting everything done in time, I quite enjoy giving these talks. We got some nice questions from the audience so overall it was a nice experience.

Often at these large conferences I feel like I’m always running around trying to catch certain talks but this year I felt they put the right sessions close to each other, which made it very easy to catch the talks I wanted to see.

As Isabelle mentioned above many groups presented long lists of variants/genes discovered by the different sequencing platforms. It felt like there is an enormous amount of data being generated but unlike GWA studies there isn’t a lot of consensus over what to do next yet. I expect next year’s meeting will focus more on this.

These are highly exciting times for Human Genetics and as a trainee in Clinical Genetics. I’m delighted to be working on the Deciphering Developmental Disorders project as part of my PhD. At ASHG I was very interested to hear about the other whole exome or whole genome sequencing projects taking place around the world. As we try to understand the variants found in individuals with developmental disorders the key messages were that we need complete phenotype information on individuals, excellent variant calls, an integrated team for analysis, and to involve experts in phenotyping.

Clinical Genetics has always been a specialty that collaborates across regions, countries and continents, more so now during this knowledge revolution we need to continue to collaborate across the world… But we first need to make sure we are speaking the same language in terms of phenotype ontology and this was discussed as being at the top of the on-going scientific ‘to do’ list.

The final plenary left me wondering how many more disease genes we will have identified at ASHG 2013? What is clear however, is that many more families in the developed world are getting important answers as to the genetic cause for their child’s difficulties, for those individuals and families in low- and middle- income countries it will take a little longer, but as we move ahead, we need to work together to help them develop their own genetics services and testing strategies.

I’ve been blogging from ASHG for four years now (from Hawai’i, Washington DC, Montreal and now San Francisco), and every year it becomes harder to come up with a theme from the meeting that distinguished it from the meetings that came before. Every year we hope that this will be the year that we will be treated to a crop of low frequency, high penetrance mutations for human complex diseases, and every year we see a little more progress but still no gold.

One thing that was clear from this meeting was how routine the mapping of associations for human traits has become. For instance, the “Chipping Away At Immune Disorders” session was a parade of new loci for a range of immune traits, mostly mapped using the Immunochip. Perhaps the most extreme example of this locus glut was Nick Erikson of 23andMe presenting, almost nonchalantly, 300 new associations to a range of human traits such as unibrow, back hair and dimpling.

However, despite how routine the association mapping stage has become, the talks did not present a unified front on what to do after you’ve discovered these associations. Some people talked about doing fine mapping, others looking across other diseases or other populations, or digging into epigenetics, model organisms, or deeply phenotyped cohorts. These are all doubtless good ideas, but are more a fragmented series of experiments than a unified approach. Now that we aren’t spending all our effort mapping associations, perhaps the field as a whole will be able to come up with consistent and shared frameworks for turning statistical association into biological insight.


  1. Darren Logan

    A few of you make interesting observations about the current lack of a framework for uncovering biological insight from association.

    I would argue that this reveals a fundamental difference between the science behind finding association and the science behind understanding function or mechanism.

    The former is now relatively standardized and therefore portable, making it feasible to report 300 associations if you have the sequence of enough people and a panel of phenotypic characteristics. (I’m not suggesting it’s trivial, but doable if you have some smart people that know what they are doing).

    Compare with this our efforts to standardize the phenotyping of 300 mutant mice. Even with that completed (taking maybe 60 people around 5 years at a cost of many millions) we don’t really have much more than a pointer towards the biological function of most mutations. For that we need to do detailed ‘secondary phenotyping’ which is a unique set of experiments and techniques for each line, often requiring the result of one before the next can be planned. Each of these is a PhD project by themselves: it’s a time consuming, expensive and risky process. Which may give one nice publication at the end… or may end up with lots and lots of confusing negative data and a submission to PLoS One.

    I think this tells us that extracting biological insight from association is extremely unpredictable and does not lend itself to a standardized approach, therefore it’s unlikely a community will converge around a unified strategy. It also means that a huge list of genotype/phenotype associations is forming, and will continue to grow, with little prospect of understanding their meaning on a similar scale or time frame.

    • Well yes I wasn’t meaning to suggest that the entirety of biology can be reduced to a standardized statistical genetics pipeline. What I was thinking was something more akin to your primary phenotyping – a standardized set of tools to statistically annotate and interpret a set of associations to generate something akin to the “pointers” you mention.

      Basically I think that we are generally not yet ready for “secondary phenotyping” of human trait associations – we could do detailed investigation on a locus-by-locus basis when we had 7 associations, but you just can’t do that for the numbers we are finding now. We need a framework to stand in between “300 associations” and “manual functional follow-up”: a set of investigations that can take us from “big list of associations” to “candidate genes, pathways and biological hypotheses that can be investigated” without requiring a PhD student per locus.

  2. Pingback: ASHG 2012 Link Roundup