On 7 September at 13:45 Katri Pärna will defend her doctoral thesis “Improving the personalized prediction of complex traits and diseases: application to type 2 diabetes”.
Professor Luca Pagani, University of Tartu and University of Padova (Italy)
Researcher Davide Marnetto, University of Turin (Italy)
Professor Harold Snieder, University of Groningen (the Netherlands)
Senior Researcher Ilja M. Nolte, University of Groningen (the Netherlands)
Associate Professor Krista Fischer, University of Tartu
Professor Reedik Mägi, University of Tartu
Dr. Ryan Daniel Hernandez, University of California (USA)
In nowadays world, common complex diseases are among the top leading causes of death globally. These diseases result from many genetic and non-genetic (e.g. lifestyle and environment) factors and from interactions between them. Since such diseases have a high health burden for the affected individual and place a heavy load on the healthcare systems, scientists are searching for solutions to delay their onset or even better, to prevent them. Evidently, differences in genetic and non-genetic components result in variation in disease risk between individuals. Therefore, prevention of such complex diseases requires a personalized approach that uses each person’s genetic and non-genetic information to predict his or her disease risk. In the current thesis, type 2 diabetes (T2D) was used as a model example of a common complex disease, T2D occurs when the blood sugar levels are too high and results in severe health complications when appropriate and timely treatment is not guaranteed. Factors such as higher age, low physical activity, high calorie intake, low socioeconomic position, smoking, and alcohol consumption have already been established as risk factors for T2D. However, the contributions of genetic risk factors and their interactions with non-genetic risk factors have not been so well explored. Therefore, the current thesis zooms in on the human genome to understand how better to use genetic information for risk prediction of T2D, leveraging on recently developed polygenic risk score (PRS – a measure combining a person’s genetic risk for a disease) approaches. Such PRSs could already enable detection of the high-risk individuals for T2D according to their genetic composition at young ages before the onset of the disease. However, there are still several limitations regarding the use of a PRS in clinical practice as its performance does not reach to the estimated levels or it cannot be constructed for each individual in a similar way due to the population-specific risk factors, causing too low estimated risks when applied in non-Europeans or admixed individuals. Therefore, current thesis presents five chapters, which mainly focus on improving the personalized prediction via genetics, tackling the current methodological limitations for PRSs, plus investigating the role of epigenetic risk factors for T2D. In the first chapter, a PRS method (called doubly-weighted GRS) was validated in two European biobanks. In the second chapter, novel PRS methods were developed to improve the PRS transferability for individuals with admixed ancestry. In the third chapter, the PRS transferability issue was investigated on a finer-scale, that is, whether a principal component projection (a method to account for population structure) could mitigate the transferability issue between two European populations. In the fourth chapter, associations of methylation scores (MSs) with prevalent T2D and its glycemic endophenotypes were tested to see whether epigenetic mechanisms could represent environmental and gene-environment effects on top of the genetics. In the fifth chapter the latest advancements in the genomics field were discussed and how to apply these in the personalized medicine framework with the prime example of the Estonian Biobank. The findings of this thesis showed that the doubly-weighted GRS indeed performed better that the traditional GRS in both European biobanks. The novel PRSs, which used the information from the method estimating genetic ancestry in a specific genetic locus could improve the prediction for the recently admixed individuals. These PRS methods made it possible to include individuals and having them benefit from personalized prediction, who were previously just excluded from the genetic studies. The traditional population-specific principal components outperformed our approach. However, the resulting PRS still contained population structure. Lastly, MSs showed a promising trend towards representing the environmental triggers for T2D and its underlying traits. In summary, the doctoral thesis resulted in more accurate and broader application of personalized prediction for complex traits and diseases leading us a step closer to personalized medicine, which makes it easier to maintain health and to prolong healthy life years.
The defence will be held in the University of Groningen (Broerstraat 5, Groningen, the Netherlands) at 12:45 (local time in the Netherlands). Online link: www.rug.nl/digitalphd.