Human Cytomegalovirus (HCMV) is found ubiquitously in humans worldwide, and once acquired, the
infection persists within the host throughout their life. Although Immunocompetent people rarely are
affected by HCMV infections, their related diseases pose a major health problem worldwide for those
with compromised or suppressed immune systems such as transplant recipients. Additionally,
congenital transmission of HCMV is the most common infectious cause of birth defects globally and
is associated with a substantial economic burden.
This thesis explores the application of statistical modelling and genomics to unpick three key areas of
interest in HCMV research. First, a comparative genomics analysis of global HCMV strains was
undertaken to delineate the molecular population structure of this highly variable virus. By including
in-house sequenced viruses of African origin and by developing a statistical framework to deconvolute
highly variable regions of the genome, novel and important insights into the co-evolution of HCMV
with its host were uncovered.
Second, a rich database relating mutations to drug sensitivity was curated for all the antiviral treated
herpesviruses. This structured information along with the development of a mutation annotation
pipeline, allowed the further development of statistical models that predict the phenotype of a virus
from its sequence. The predictive power of these models was validated for HSV1 by using external
unseen mutation data provided in collaboration with the UK Health Security Agency.
Finally, a nonlinear mixed effects model, expanded to account for Ganciclovir pharmacokinetics and
pharmacodynamics, was developed by making use of rich temporal HCMV viral load data. This model
allowed the estimation of the impact of immune-clearance versus antiviral inhibition in controlling
HCMV lytic replication in already established infections post-haematopoietic stem cell transplant.