Care, Angharad 
ORCID: 0000-0003-2131-0406
(2021)
Using “Omics” to Discover Predictive Biomarkers in Women at High Risk of Spontaneous Preterm Birth.
PhD thesis, University of Liverpool.
|
Text
201149627_Oct2019.pdf - Unspecified Download (8MB) | Preview |
Abstract
Spontaneous preterm birth (sPTB) is a complex pregnancy syndrome that remains poorly understood and is associated with significant perinatal morbidity and mortality worldwide. Current research suggests that there are multiple disordered physiological processes that trigger a final common pathway of early labour, rather than a single specific cause. It is this heterogeneity that has hindered the discovery of a single predictive biomarker and existing screening methods for sPTB prediction are insufficient to detect all women at risk. Consequently, our inability to identify women at risk inhibits efforts of prevention, which cannot be achieved without better understanding of causation or a more robust way of accurately discriminating those at high risk. The development in “omics” technology has led to exciting breakthroughs in other areas of medicine and offers new avenues of investigation for sPTB prediction. The primary aim of the thesis was to establish a way of combining different types of ‘omics’ analysis from the same individual in a pilot study to identify candidate biomarker predictors or pathways. Three different “omic” methodologies; genomics, transcriptomics and metabolomics, were used to analyse blood taken from asymptomatic women high-risk for sPTB at 16 and 20 weeks of pregnancy. Lastly, I investigated if there are distinct differences in biomarkers between PPROM and sPTB subgroups of spontaneous preterm birth. On an individual omics level only transcriptomics showed an association with sPTB. Gene set enrichment in this population demonstrates that the selenoamino acid pathway differentiates asymptomatic high-risk women. Hierarchical clustering in a non-linear distance matrix differentiated all but one of the sPTB and PPROM cases. More studies are required to validate the findings from our analysis. Data from each omics discipline was combined together in a single data matrix and machine learning analyses applied. The area under the curve (AUC) of receiver operating characteristic (ROC) values for Linear discriminant analysis (0.90), Genetic expression programming (0.70), K-Means (1.00), Linear support vector machine (0.96), Support vector machine with a Gaussian Kernel (0.96), Probabilistic neural network (1.00) and Random Forest (0.96) demonstrate that most machine learning methods perform well on our dataset. Sample sizes needed to reach excellent (AUC = 0.9) vs. moderate (AUC = 0.7) prediction performance were found to be within realistic ranges. This study provides a conceptual analytical framework for the prediction of sPTB. For a larger cohort prediction power is excellent, making individualized preterm prediction a realistic possibility.
| Item Type: | Thesis (PhD) |
|---|---|
| Divisions: | Faculty of Health and Life Sciences > Institute of Life Courses and Medical Sciences > School of Medicine |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 26 Mar 2021 09:36 |
| Last Modified: | 18 Jan 2023 22:58 |
| DOI: | 10.17638/03116086 |
| Supervisors: |
|
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3116086 |
Dimensions
Dimensions
