Use this URL to cite or link to this record in EThOS:
Title: Finite mixture models in statistical genetics with applications to schizophrenia and motor neurone disease
Author: Gillett, Alexandra Claire
ISNI:       0000 0004 7660 1523
Awarding Body: King's College London
Current Institution: King's College London (University of London)
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Genetic and environmental risk factors act together in common, complex disorders to trigger disease onset. The genetic architecture of disease can range from Mendelian to polygenic, and environmental variables can exert considerable influence on disease risk, both independently and as modifiers of genetic risk factors. Many different methods have been developed to predict disease risk from genetic and environmental risk factors, but given the complexity of disease, it is difficult obtain predictions of risk that have clinical utility. Family history of disease captures information about genetic risk variables, and environmental risk factors that are correlated within families. For this reason, family history can be used in combination with identified genetic and environmental risk profiles to improve prediction accuracy. This PhD develops and applies methods for genetic risk prediction using family history. We assume that the true model of disease is a liability threshold mixture model (LTMM), also known as a mixed inheritance model. This finite mixture model can include large effect genetic risk loci and a polygenic component. We start by presenting a simple approach to estimate the variability in liability to disease attributable to a polygenic component, known as a polygenic risk score (PRS). We call this quantity the heritability of the PRS. The proposed method does not require raw data; only summary statistics, and under certain assumptions provides an unbiased estimator for the PRS heritability. Contained within this approach is a method to transform univariate effect size estimates from the logit scale to the liability scale, using the normal and logistic cumulative distribution functions. We then use the LTMM to derive equations for disease risk given a PRS, a set of major genetic loci, a group of environmental risk factors and family history of disease, using summary statistics available in the literature. Alternative risk equations, derived using the log-linear model, are also presented. The utility of these equations are demonstrated by estimating the risk of schizophrenia. Finally, we apply a finite mixture model in latent class cluster analysis to identify subgroups of amyotrophic lateral sclerosis (ALS) patients using clinical phenotype data. ALS displays substantial clinical heterogeneity, and stratifying cases into homogeneous subgroups may be a powerful tool to identify subtype-specific risk loci in genetic studies.
Supervisor: Lewis, Cathryn Mair ; Al-Chalabi, Ammar Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available