Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Identifying immunological biomarkers of sepsis using cytometry bioinformatics and machine learning

Burton, Ross Jake 2022. Identifying immunological biomarkers of sepsis using cytometry bioinformatics and machine learning. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of 2023BurtonR PhD.pdf]
PDF - Accepted Post-Print Version
Download (65MB) | Preview
[thumbnail of Cardiff University Electronic Publication Form] PDF (Cardiff University Electronic Publication Form) - Supplemental Material
Restricted to Repository staff only

Download (119kB)


Sepsis is a leading cause of mortality and significantly strains healthcare systems worldwide. Improving sepsis care and outcomes depends on appropriate risk stratification and timely identification of the causative pathogen to guide patient management and treatment. Enormous efforts have been made to identify diagnostic and prognostic biomarkers to aid decision making, but to date, they have failed to identify candidates with acceptable accuracy and precision to have an impact in the clinic. Past studies have often focused on individual biomarkers without considering the potential benefit of multi-marker panels incorporating deep immunological phenotyping. This work addressed this issue with a cross-disciplinary approach that integrated sepsis biomarker discovery, cytometry bioinformatics, and supervised machine learning. Firstly, a novel framework for cytometry data analysis was developed, along with a new ensemble clustering algorithm that reduced the risk of biasing exploratory analyses with the application of a single clustering technique. Secondly, the analysis framework was applied to a study cohort of severe sepsis patients, and their early immunological profile consisting of cellular and humoral parameters (within 36 hours of diagnosis) was determined. The captured immunological parameters were then combined with routine clinical data and lipid plasma concentrations to generate interpretable machine learning models for predicting mortality and the underlying cause of infection. The generated models discriminated between survivors and non-survivors, and between Gram-negative and Gram-positive infections, and identified potential combinations of biomarkers with predictive value.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Medicine
Uncontrolled Keywords: Sepsis - Biomarkers - Unconventional T cells - Clustering - Cytometry - Bioinformatics - Machine learning
Date of First Compliant Deposit: 21 February 2023
Last Modified: 21 Feb 2023 15:15

Actions (repository staff only)

Edit Item Edit Item


Downloads per month over past year

View more statistics