Exploration of machine learning approaches with genome-scale metabolic model-generated fluxes

  • Giuseppe Magazzu

Student thesis: Doctoral Thesis


Biology, from the ancient Greek, “study of life”, is the most fascinating yet complex of all sciences.
Its understanding is paramount in solving many current problems we face in our society, from curing
seriously debilitating diseases, to safely devising new drugs, to determining guidelines for disease prevention.
However, its complexity has so far hindered medical progress, and the objective impossibility
of perfectly replicating experiments has contributed to less robust results and many non-reproducible
claims. Lately, the advancement of computing technologies has led to the development of mathematical
and computational methods which could shed light on the mechanisms of life which are still obscure
to us to date. In particular, progress in Systems biology has opened the way to new techniques to
be employed to solve long-standing problems in more efficient and robust ways. One of the newly
designed methods, genome-scale metabolic models, can help simulate metabolic conditions and draw
links between the molecular transformations happening at a small-scale level and the systemic modifications
that organisms experience. In this work, we have explored the usefulness of such models and
investigated possible approaches in the form of case studies, with the adoption of machine learning
techniques. These techniques, which aim at discovering invisible patterns in the data the significance
of which could lead to fundamental discoveries or directions for future medical research, currently
represent the state-of-the-art approaches in countless of modern applications, and serve as the first
choice when trying to advance in cutting-edge research scenarios. Our work demonstrates that some
scope for these models exists, in particular in the field of precision medicine.
Date of Award2023
Original languageEnglish
Awarding Institution
  • Teesside University
SupervisorZulfiqur Ali (Supervisor) & Claudio Angione (Supervisor)

Cite this