Study Watch : Common genetic variants on 5p14.1 associate with autism spectrum disorders

There were two studies published at the end of last month with a large amount of fanfare announcing that a gene was discovered that might be responsible for as many as 15 percent of cases of autism.

I wanted to write about them earlier but the studies are complex and understanding what exactly they are saying has taken me some time.  I am going to start with the main study and should follow up with the second study shortly.  

The main study was published on April 28th in Nature -

Common genetic variants on 5p14.1 associate with autism spectrum disorders

This study has an large list of authors with over 50 people attaching their name to the study - I don't recall ever seeing a study with so many names attached to it.  The study itself is a very dense five pages of material although the supplementary material weights in 65 pages.

In this study the authors were attempting to identify common genetic risk factors for autism.  They first started by looking for associations in a data set from the Autism Genetic Resource Exchange (AGRE) that ultimately contained 3,101 subjects of European descent from 780 families.    

There were no associations found that were significant.  However, the authors proposed that there could be meaningful associations that among the results that showed some possible association yet did not rise to the level statistical significance.

To that end examined a second data set from an Autism Case-Control cohort (ACC) containing 1,204 case subjects and 6,491 control subjects that were provided from multiple projects across the country and from Childrens Hospital of Philadelphia.

No significant associations were found in this data set either.

However, the authors took it a step further and did a meta-analysis of the two data sets together using very recent techniques and found a result that reached significance and several others that came close.  The most significant association was found at 5p14.1 with the others being located at several nearby locations.

To confirm these findings two other data sets were examined. The first was the Collaborative Autism Project (CAP) cohort with 504 subjects and the second was the Center for Autism Research and Treatment (CART) cohort with 108 case subjects and 540 controls. 

Neither of these two data sets yielded a significant association on their own but like the AGRE dataset they showed some associations that did not reach the level of significance.  However, when these data sets were combined with the original two the areas identified above were again significant.

All of the areas of interest identified are location in a region between the CDH9 and CDH10 genes.  The authors hypothesized that the locations identified were showing functional variations in one of two two genes.

These genes and others that could be related are thought to be involved play a role in shaping the physical structure and functional connectivity of the brain.  These are two areas that other research as identified as potential issues in people with autism, so the results do fit with other theories of autism. 

As an aside, there is a post at autism.about.com where Lisa Jo Rudy published some questions and answers from the study's lead author that is worth reading.

I am not sure what to make of this study.

It is not straightforward at all and the amount of information published with the study is far larger than is normal.  I think this is likely because the authors wanted to cover all of their bases yet I have to wonder why those chose to add all of it.  

I am confused as to why a significant association was not present in any of the individual data sets but only appeared when the data sets were analyzed together.  I would think that if the claim of 15% of autism cases contained in the press releases for the study were correct that the association would have been more apparent from looking at any of the data sets individually.  Or I would have thought that it would have been identified in one of the other large genetic searches that were done prior to this.

I also have to wonder at the methodology used to perform the meta analysis of the data sets.  From what I can gather the technique is very new - the reference that the study provides to the methodology was published just a month or two (October 2008) before this study was submitted to the publisher (Nov 2008).  I did not investigate the methods in depth but from what little I did see gives me pause.  

The results of the study rest on the meta analysis of the data sets - if this methodology is flawed or was implemented badly then association found would fall apart. 

So all in all I would think that this is a potentially promising study but that further confirmation is required.