2009 © Institute for Molecular Medicine Finland FIMM
FiDiPro Professor Joseph D. Terwilliger, PhD, FIMM and Columbia University, USA
As a statistical geneticist, the main foci of my research have been development and implementation of statistical methods for inference about human genetics, developing and applying novel study designs using natural experiments to improve the power of human genetic investigations, and organizing and teaching workshops both in Finland and throughout the developing world on “logical reasoning in human genetics”.
The primary project I have been involved at FIMM has been the development and implementation of likelihood-based algorithms for inference about linkage and linkage disequilibrium jointly on familial data. While much emphasis in the field has been on the analysis of unrelated individuals in case-control and cohort settings, our emphasis has been based on approaches to analyzing genome-wide SNP and sequence data in collections of families. The primary advantage of the Finnish population in human genetic research has been the ability to construct and ascertain familial data through the use of the well-established registries of medical diagnoses and family structures. In recent years much effort has been invested in attempting to do genome-wide association studies in Finland as elsewhere using “unrelated” subjects, with very little success. The field as a whole is beginning to realize that there are likely to be a significant number of rare variants underlying the risk of most common diseases, and for addressing this problem, a return to analysis of family data has been widely recognized as critically important. Methods for analyzing this sort of data have been lacking because very few studies in recent years have applied the new technologies to large families.
Another aim of my research group has been to explore the likely genetic architecture of complex traits through forward evolutionary simulations under user-specified models of phenotype-based natural selection, mutation, recombination and demographic history. Unlike most existing approaches to this problem, we simulate individuals and their reproduction forward in time over tens of thousands of generations, generate new mutations with simulated effect sizes, and allow natural selection to occur based on phenotypes simulated for each individual in each generation in a biologically natural way. We hope to generate more reasonable hypotheses about the genetic architecture of natural phenotypic variation, now that we have know that there are not many common genetic variants of large effect, as many researchers had been hoping. We hope to thus better understand why gene mapping in complex traits by genome-wide association analysis has been such an overwhelming failure.
Ultimately it is not the statistical analysis method that determines the success or failure of a gene mapping study, but rather the underlying biological truth and the study design used to query the biological reality. Finnish genetics has been successful over the years because of its emphasis on the use of the natural history of the Finnish population as a natural experiment from which interesting questions could be asked that were impossible in other populations. However, there are many other natural experiments around the world that can be used in an analogous way to answer different questions. Much of my work has involved working with populations in the developing world where many such “natural experiments” can be found, to help the local researchers use their populations in unique and creative ways to ask questions about human biology. My work in this area at the moment is concentrated on populations in Kazakhstan, Korea and Venezuela, where I am involved in a number of collaborative research projects, to look at interactions between genetic and enviro-cultural factors on normal human variation and health.