Polygenic score application to complex trait prediction

Student thesis: Doctoral ThesisDoctor of Philosophy


Decades of quantitative genetics research has led to the conclusion that all human behavioural traits tested to date show genetic influence to varying degrees. However, powerful individual-level genetic prediction based on measured genetic variation has only become feasible more recently through technological and methodological advancements. Especially over the past decade, the application of genetic prediction methods has grown exponentially, permeating research in the social, behavioural and biomedical sciences. 
This thesis seeks to investigate developmental and multivariate research questions, as well as genotype-environment interplay through the use of a broad range of polygenic scores analysis approaches and phenotypes. The research sample in this work was the UK Twins Early Development Study (TEDS), utilised to investigate the predictive accuracy of a genome-wide polygenic score (GPS) for years of education for a variety of cognitive and noncognitive traits (Chapter 2 and 3); the shared genetic aetiology between a range of psychiatric disorders (Chapter 4); potential genotype-environment interactions (GxE) for education and intelligence using the polygenic score for years of education (Chapter 2); the influence of evocative genotype-environment correlation (rGE) in the relationship between child bodymass index and parental feeding practices (Chapter 5); and finally, the effect of passive rGE on polygenic score prediction estimates across various trait domains by comparing within- and between-family polygenic score predictions (Chapter 6). 
This thesis provided evidence that 1) target trait prediction estimates of the GPS for years of education represent the strongest polygenic prediction of any behavioural trait; 2) the GPS for years of education is associated with a wide range of traits, including cognition, personality, BMI, physical and mental health, and environmental measures; 3) the substantial genetic overlap between psychiatric disorders may be due to a common genetic factor; 4) there is no evidence for GxE for cognitive and educational outcomes using the GPS for years of education; and 5) evocative and passive rGE can be detected and quantified through GPS analysis. 
Leveraging a broad range of phenotypes and GPS, this thesis illustrates the usefulness of the polygenic score approach in research. General implications, limitations and futuredirections will be discussed (Chapter 7).
Date of Award1 Oct 2019
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorRobert Plomin (Supervisor), Jonathan Coleman (Supervisor) & Paul O'Reilly (Supervisor)

Cite this