A fascinating new study [press release here, full text freely available here] out of the Institute of Psychiatry at King's College London shows that the relative contributions of genetic and environmental influences on various cognitive and behavioral traits differ by location. In other words, if you live in location A, trait X might mostly depend on your genetic makeup, whereas if you live in location B, trait X might be more susceptible to variation in your environment regardless of your genes. Pretty neat stuff!
The authors assessed 6,579 pairs of 12-year-old twins in the UK on 45 different characteristics. Monozygotic ("identical") twins share all of their genes and dizygotic ("fraternal") twins share, on average, half of their segregating genes. There are components of the environment that are shared by both twins and components that are unique to each twin. By using these different correlations, we are able to determine how strongly each part of the equation affects a given trait by seeing how much more similar monozygotic twins are on that trait than dizygotic twins.
Although most researchers know better, we tend to talk about these estimates of genetic and environmental influence as if they are static and fixed. This study serves as a great reminder that this is not the case. A trait can appear more genetic or more environmental depending on the genetic makeup of the sample used or the environment the sample lives in.
The authors focused on just one of the 45 traits that they collected for this paper: a composite of teacher-rated classroom behavior problems. I've copied a portion of their figure below. It shows the strength of the influence of the non-shared environment on classroom behavior problems. The warmer (pinker) colors represent a stronger environmental effect and the cooler (bluer) colors represent a weaker environmental effect. As you can see, there is a trend toward Londoners' classroom behavior problems being more affected by the environment than those in outlying areas.
|Figure 3a. Unique environmental influences on teacher-rated classroom behavior problems|
So what is it about London that makes the environment more potent for this particular trait? The authors posited it could be income variability. In other words, London has more people who are very rich and more who are very poor, while the rest of the UK is more evenly distributed. An environment that is more variable and extreme can have more of an effect than an environment that doesn't change much from person to person. When they plotted income variability, the resulting map looks very similar to the one above. This correlation doesn't prove that income inequality itself causes classroom behavior problems. It just provides a hint about what other mediating variables to look into more closely in the future.
You can download the files and scripts to explore the rest of the rich dataset yourself here. It's free, very quick and easy, and there's even a video to show you how to manipulate the program. I played around with it a little bit tonight to check out some of the autism data. The results of my brief exploration are below.
They administered the Child Asperger Syndrome Test to the twins' parents and teachers. In addition to an overall total, the measure has subscales for social, nonsocial, and communication. There are maps for the genetic component, common environment, unique environment, and overall variation. All of that means that there are 32 maps for autism alone! That's a lot to digest. I quickly realized that I could easily get sucked down a long rabbit hole of analysis and interpretation, so I had to limit myself (for now, at least). I looked at the parent-reported social subscale and overall autism score for each component. You can see my great image manipulation skills below.
|My quick maps for some of the ASD data|
We already know that income variability is one factor that could potentially mediate this relationship. In general, it seems that the relative contributions of genetic and environmental influences on the development of autism differ by something that differentiates the UK's urban and rural areas. This is an exciting avenue for future research.
In addition to the important findings and implications of the data themselves, the authors are hopeful that this paper will help propel an already-developing "trend towards integrating visualization into the analytic process, instead of approaching it as a way to effectively communicate the outcome of a completed study." In other words, we scientists should take advantage of our ability to spot patterns by making various sorts of images of our data for our own exploratory purposes. Don't wait to make a pretty graphic until it's time to publish!
In order for this to happen, we need to forge collaborations with people who specialize in this sort of thing. I'm not an expert in generating effective visualizations and I'm never going to be the best one to write the code for it, so apparently I need to start making some new friends... [edit: ew I accidentally used the wrong too before :( ]
O S P Davis, C M A Haworth, C M Lewis, & R Plomin (2012). Visual analysis of geocoded twin data puts nature and nurture on the map Molecular Psychiatry DOI: 10.1038/mp.2012.68