Computational biology applications in the study of complex systems

Bravo Merodio, Laura ORCID: 0000-0001-8878-8434 (2023). Computational biology applications in the study of complex systems. University of Birmingham. Ph.D.

Preview

BravoMerodio2023PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (13MB) | Preview

Abstract

In biomedicine, the advent of digitalization and big improvements in computing power and high throughput technologies has yielded an unprecedented amount of data. To harness this data’s full potential, research has become increasingly computational, with core tools of data science such as machine learning required to decipher patterns, help extract meaning and uncover underlying connections. In this work, we have explored key leading areas of computational research, from discovery science to translational medicine and complex system studies. First, a machine learning pipeline was developed to inform bioinformatic research. After validation, it was applied on a novel dataset generating insight onto the predictive power of immune features in ultra-early trauma injury, generating leads of possible biomarkers associated with the development of MultiOrgan Dysfunction. Then, in order to make our pipeline accessible, an interactive, simple, free and open-source supervised machine learning webtool was developed. Adapting this same framework to the realm of clinical translation, we then deployed two prognostic models for decision support in surgeries during the COVID-19 pandemic. This was done in collaboration with the NIHR surgical team. Lastly, we explored systems biology approaches by studying the complexity behind ageing and health to disease transitions. For this, we leveraged the UK Biobank data as a rich source of deeply phenotyped information. By calculating biological age (PhenoAge) longitudinally for nearly 400,000 participants, we identified four distinct ageing categories ranging from healthy to unhealthy ageing. These different trajectories were characterised by their chronic diseases and genetic makeup, revealing a strong association of metabolic dysfunction with unhealthier phenotypes and immune-related signals for healthier ones. Also, strong opposite-effect associations of longevity-related variants were found, with novel regulatory elements postulated as possible drivers of unhealthy phenotypes, opening new avenues for future study. Overall, our findings highlight the crucial role of computational methods in biomedicine and their potential to transform clinical practice.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):