Renowned expert in the epidemiology and statistics of genetics, dr. Peter Visscher will make a total of three visits to the VU/ CNCR. Visscher received a TOP visiting professorship grant, subsidised by the Royal Netherlands Academy of Arts and Sciences (KNAW).
Getting more out of even more data
An expert in the epidemiology and statistics of genetics, Peter Visscher will make a total of three visits to the VU CNCR, working on complex data analysis for the neurosciences. Visscher: ‘We just have to be smarter in our application of our methodology. And as always in science asking the right questions is a good starting point.’
’ Do not underestimate the amount of data that is being generated at this moment, it is just mind-boggling! And this is going to increase exponentially.’ Peter Visscher works as a Senior Principal Research Fellow at the Queensland Institute of Medical Research in Australia, and holds a professorship at the University of Queensland, the Queensland University of Technology and Griffith University. This year he will make a series of trips to Amsterdam on a TOP visiting professorship grant, subsidised by the Royal Netherlands Academy of Arts and Sciences (KNAW). Together with his host Danielle Posthuma from the Centre for Neurogenomics and Cognitive Research, he will focus on their shared fascination: improving the methodology of data mining for the huge amount of genomic data.
‘The current technology in genetic research is to measure DNA markers on gene chips. The state of the art chips now scan a million single nucleotide polymorphisms (SNPs) at the same time. New technology is moving towards resequencing the whole genome of individuals. Compare that to the whole Human Genome project that took twenty years and billions of dollars: we are now getting towards a thousand dollars per genome, which is an amazing scale-up of technology. Our common interest is trying to make sense of the very large amounts of genotype data that has become available. Therefore instead of setting up new experiments Danielle and I will be working on methodology, especially on complex data analysis.’
Peter Visscher explains that Genome Wide Association (GWA) studies helped to identify genetic variants and genes that modify the risk of developing a certain condition, but that the effect sizes generally are quite small. ‘If you have such a variant gene, your risk is increased, but often the difference is really small. To assess these subtle effect sizes we need very large data samples, for example the DNA and datasets collected in twin registries.
’The question is: if the effect sizes are small, and single genetic variants give only a small increase in risk, what can this tell us about the biology? That is where neuroscience comes in. I have a longstanding interest in traits that affect psychiatric disease, especially schizophrenia and bipolar disease. Here at the VU CNCR there is a very strong group on the biology ofneurogenetic disease.’
There are many blank spots in the understanding of how genetic vulnerabilities relate to psychiatric diseases, but using innovative data analysis approaches Visscher hopes to fill in some of these blanks. ‘We take an unbiased approach. Danielle and I try to analyse many of these genetic variants at the same time, rather than just looking at them one by one. Even if the effect sizes are small this enables you to identify new biological pathways that could be involved, and you immediately might start to think of a drug that would intervene in that gene and the pathway. Intervening in a pathway often means you do not need to have a gene with a huge effect for it to be biologically relevant.’
The sheer amount of data forces Visscher to rethink the statistics that is used to test the significance of his findings, he explains. ‘We have to deal with very large datasets, and multiple levels of data, which is not just restricted to genotype data. You could measure transcription, the proteome, methylation: the real challenge is to bring these datasets together, and come up with strong hypotheses that can be tested in a different data set, or make a prediction about how things work and then test that prediction. 
The amount of data is so huge, that you can formulate so many hypotheses that one of them seems to be statistically significant, but maybe you just tested too many of them and one ended up to be positive. It is very easy to be misled by what seems to be a signal, but is just something that you see because you have been looking so hard. We have to set the bar extremely high for statistical significance in these experiments.’
Three times two weeks Peter Visscher will spend in Amsterdam this year, and his agenda is packed. ’ One aim is to write some scientific papers together with Danielle Posthuma, which we already started. Another goal is to try out novel network analyses, using existing data from my group in Australia, from the public domain and from Amsterdam. And last but not least, we want to do a grant application together to restart our successful series of teaching workshops on complex genetic data analysis, which we jointly organised some years ago.’

By Lucas Maillette de Buy Wenniger