2021eResearchReport - Flipbook - Page 23
Using machine learning, we identified
specific functional categories that were
unique to, and common among, nonLactobacillus-dominated (high-diversity)
CVM, BV, and HR-HPV infection. Such
differentially-abundant functional
categories may represent important
diagnostic biomarkers of BV and HR-HPV
infection. Shared functional features
suggest interconnectivity between highdiversity CVM and these reproductive
outcomes. Our findings are hypothesisgenerating and require proof-of-concept
functional studies to examine the
relevance of these potential biomarkers in
women’s reproductive health and disease.
Onywera H, Anejo-Okopi J, Mwapagha LM, Okendo
J, Williamson AL. (2021 Jun). Predictive functional
analysis reveals inferred features unique to
Below: Dr Gregory Distiller applied the
Spatial Capture-Recapture model to model
the animal density of jaguars in Belize.
Image credit: Panthera Belize
22 eResearch Report | 2019-2020
cervicovaginal microbiota of African women
with bacterial vaginosis and high-risk human
papillomavirus infection. doi.org/10.1371/journal.
pone.0253218
Understanding animal behaviour with HPC
Quantifying the distribution of daily
activity is an important component of
behavioural ecology. Historically it has
been difficult to obtain data on activity
patterns, especially for elusive species.
However, the development of affordable
camera traps and their widespread usage,
has led to an explosion of available data
from which activity patterns can be
estimated.
Spatial Capture-Recapture (SCR)
models have become the standard
approach used to model animal density.
The data for these models is typically
broken up into discreet occasions but a
recently developed Continuous-Time (CT)
framework for SCR models utilises the
actual time of capture, which is easily
extracted from camera-trap surveys. In
addition to estimating density, CT SCR
models estimate expected encounters
through time. Cyclic splines can be used
to allow flexible shapes for modelling
cyclical activity patterns and hence, these
models can estimate how detectability
changes through a 24-hour daily cycle.
Furthermore, the fact that the detection
function in SCR models also incorporates
distance means that space-time
interactions can be explored.
The method is applied to a dataset on
jaguars and demonstrates how CT SCR
models can be used to explore hypotheses
about animal behaviour within a formal
modelling framework. While SCR models
were developed primarily to estimate and
model density, the models can be used
to explore processes that interact across
space and time, especially when using
the CT SCR framework that models the
temporal dimension at a finer resolution.
Distiller G, Borchers D, Foster R, Harmsen B.
(2020 Oct). Using Continuous‐Time Spatial
Capture–Recapture models to make inference
about animal activity patterns. doi.org/10.1002/
ece3.6822
HPC and vaccine development.
The Cryptococcus neoformans fungus
causes disease in immunocompromised
patients such as transplant recipients.
This fungus is enclosed in a protective
slimy capsule that protects the fungus
from attack by the host’s immune
system. The capsule is mostly composed
of a polysaccharide molecule called
GlucuronoXyloMannan, or GXM. This
GXM polysaccharide varies in composition
according to the strain of C. neoformans.
GXM is a potential target for an anti-C.
neoformans vaccine. However, the
shape and the key features of the GXM
molecule are not known. This is important
information for the development of an
effective vaccine. In this paper, molecular
modelling shows that the long GXM
molecule is extended, quite inflexible and
decorated with fringes of side chains .
This shape explains some puzzling clinical
observations and overall, the project
demonstrates that molecular modelling
can play a useful role in the rational
design of conjugate vaccines.
Kuttel M, Casadevall A, Oscarson S. (2020
Jun 04). Cryptococcus neoformans Capsular
GXM Conformation and Epitope Presentation:
A Molecular Modelling Study. doi.org/10.3390/
molecules25112651
eResearch, HPC and pharmacometrics
Pharmacometrics is the science of
developing and applying mathematical
and statistical methods to characterise,
understand and predict drug levels
and effects on the human body. The
major advantage of pharmacometrics is
the ability to pool data from different
studies and translate drug behaviour
across different populations. Coupled
with High Performance Computing
(HPC), pharmacometrics plays a pivotal
role in data-driven clinical research in
countries with limited resources. In 2018,
over 10 million people were diagnosed
with tuberculosis and 1.4 million died.
Tuberculosis is the leading HIV-associated
opportunistic infection and the main
cause of death, particularly in resourcelimited countries. Sub-Saharan Africa
bears the highest burden of HIV with
approximately 20.7 million people
living with HIV as of 2019. The UCT
pharmacometrics research group supports
the application of pharmacometrics
to investigate the pharmacology of
antiretroviral, antituberculosis and
antimalarial agents with the aim of
improving the way we treat patients for
these infectious diseases and optimising
dosing in neglected populations (such as
children and pregnant women) in African
countries. We aim, with these state-ofthe-art methodologies, to continually
improve patients’ care and health.
Ignatius E, Abdelwahab M, Hendricks B, Gupte
N, Narunsky K, Wiesner L, Barnes G, Dawson
R, Dooley K, Denti P. (2021 Jan). Pretomanid
Pharmacokinetics in the Presence of Rifamycins:
Interim Results from a Randomized Trial among
Patients with Tuberculosis. doi.org/10.1128/
AAC.01196-20
Supporting research 23
