Friday, May 18, 2012

IMFAR 2012: Genetic Variants in ASD

Any omissions or errors in this summary are on TPGA; we have tried to include explanatory links for specific scientific terms. -SR

Common and Genetic Variants in the Etiology of ASD: Where is the Field Heading?

Bernie Devlin
University of Pittsburgh
Statistical geneticist, helps design projects & studies.

Dr. Devlin's goal: convince you that we know quite a bit about the genetics of autism, in a few years we'll know more.

Exciting times! Are those times going to continue? Devlin maintains, yes.
  •     Using array-based tech & high throughput sequencing,
  •     Samples and collaboration
He considers that discovery of autism risk genes is going to increase exponentially in near future.

How will we get there?
  • Autism Genome Project (AGP)
    • Trios: Mother, Father, ASD child
    • Families could be multiplex (2+ ASD individuals per family) or simplex (1 ASD individual per family)
  •     Simons Simplex Collection (SSC)
    • Families are mostly quartets (Mother, Father, ASD individual, sibling)
    • Family fully phenotyped
    • Non-ASD sibling fully characterized for comparison

SSC CNV (Copy Number Variation) Study
  • Performed same analysis as AGP, but with SSC data, including de novo events.
  • Conclusions:
    • Most CNVs not relevant for ASD risk
    • More de novo events are related to ASD but not all

Many of our discoveries involve rare events. Rareness challenged statistical inference, and really depends if they arise de novo or not.

Gene Richness: Are they hitting a single gene in that region? Or are they hitting multiple autism risk genes in a single CNV? If so, how can we find them?
  • Smaller events
  • Gene expression, how altered by CNVs
  • Sequencing analyses, look for differences re: controls
All three could work!

Overall results:
  • Recurrent gene disruptions identify ASD genes
  • Older fathers throw more mutations, older moms contribute too

Negative impact of damaging CNVs on rate of de novo SNVs
  • We need to capture more genetic variation in the genome if we're going to understand this better.

What about standing variation (SNV)? (His term for inherited genetic variation, de novo or otherwise, that was undetected)
  • The signal for risk genes is not so obvious.
  • Recessive/Compound inheritance is a bit more promising, but signals still not outstanding
  • Similar in pattern to CNV results -- but still insufficient to understand ASD risk genes completely
  • Useful complement to de novo studies

Why de novo?
  • When you look at them as CNVs or SNVs, do a better job of producing more ASD risk genes
  • Even though rare, have much better signal to noise ratio
    • Half of all de novo SNVs are hitting autism risk genes
    • Half of all gene rich de novo CNVs also contain autism risk genes
Autism Sequencing Consortium
  • Goal: collaboration to analyze worldwide sequence data for ASD gene discovery
  • 7500 families + >1500 probands/controls

The key piece of info is that for every two nonsense events that fall on the same gene in autism probands, one is an autism risk gene.  There's a heck of a lot of info there. If we're smart about it and pool our data and analyze it with emerging technology, we'll identify the autism risk genes more specifically.

Autism is binary (you have it or don't) from a diagnostic space, but geneticists see it as a continuum.

  •  AGP estimates that heritability of ASD is ~60% for multiplex families, ~40% for simplex families.

We actually know a great deal about the genetics of autism, and within a few short years we'll know a lot more. We'll have a fairly large yield of autism risk genes within the next few years, which will yield sensible drug treatment targets.

Momentum for discovery of autism risk genes is huge and due to:
  •     Pooling data
  •     Funding   
In five years, research into autism risk genes even become passe because we'll know so much. Translation will be the key for ASD in the near future

Footnote: Devlin thought important to state that the notion Genes or Environment is outdated -- we need to consider Genes AND Environment.