Using big data for mathematical models of the human genome - implications for psychiatric genetics

Genetics is an area of large opportunities for machine learning. The largest genome-wide association studies (GWAS) have already exceeded one million individuals and contain information on tens of millions of genetic variants, jointly estimated to account for up to 80$\%$ of the variability in complex human traits, including psychiatric disorders. Nevertheless, translating this knowledge into a clinical application remains challenging, and the practice of utilizing individual genetic information to predict disease has been judged to provide little to no useful information. In this seminar we will highlight recent successes of the largest GWAS studies and discuss the limitations that hinder an effective application of machine learning techniques in human genetics. You will learn about the statistical methodology behind these studies, including the Bayesian mixed-model analysis and Restricted Maximum Likelihood. You will also learn about polygenic risk scoring and precision medicine, which is already effective in personalized risk prediction for certain types of cancer and risk stratification for Alzheimer's disorder.
Speakers: Oleksandr Frei and Kevin O`Connell, post-doctoral researchers from the Norwegian Centre for Mental Disorders Research (NORMENT), Oslo, Norway.
Speaker info: - Kevin O`Connell has a PhD from the University of Cape Town, South Africa (2014). Since 2018 he has worked as a post-doctoral researcher on human genetics at NORMENT. Prior to this he worked as a post-doctoral researcher at Stellenbosch University, South Africa, on the genetics of antipsychotic treatment response in schizophrenia patients. - Oleksandr Frei has a PhD from the Moscow Institute of Physics and Technology (2013). Since 2016 he works as a post-doctoral researcher on human genetics at NORMENT. Prior to this he was involved with different applications of machine learning, including Churn prediction (Forecsys), Search in SharePoint and Exchange (Microsoft), Geology and Modeling (Schlumberger), and Topic Modeling (BigARTM).