dc.description.abstract
[eng] The vast and continuously increasing volume of available biomedical data produced during the last decades opens new opportunities for large-scale modeling of disease biology, facilitating a more comprehensive and integrative understanding of its processes. Nevertheless, this type of modelling requires highly efficient computational systems capable of dealing with such levels of data volumes.
Computational approximations commonly used in machine learning and data analysis, namely dimensionality reduction and network-based approaches, have been developed with the goal of effectively integrating biomedical data. Among these methods, network-based machine learning stands out due to its major advantage in terms of biomedical interpretability. These methodologies provide a highly intuitive framework for the integration and modelling of biological processes.
This PhD thesis aims to explore the potential of integration of complementary available biomedical knowledge with patient-specific data to provide novel computational approaches to solve biomedical scenarios characterized by data scarcity. The primary focus is on studying how high-order graph analysis (i.e., community detection in multiplex and multilayer networks) may help elucidate the interplay of different types of data in contexts where statistical power is heavily impacted by small sample sizes, such as rare diseases and precision oncology.
The central focus of this thesis is to illustrate how network biology, among the several data integration approaches with the potential to achieve this task, can play a pivotal role in addressing this challenge provided its advantages in molecular interpretability. Through its insights and methodologies, it introduces how network biology, and in particular, models based on multilayer networks, facilitates bringing the vision of precision medicine to these complex scenarios, providing a natural approach for the discovery of new biomedical relationships that overcomes the difficulties for the study of cohorts presenting limited sample sizes (data-scarce scenarios).
Delving into the potential of current artificial intelligence (AI) and network biology applications to address data granularity issues in the precision medicine field, this PhD thesis presents pivotal research works, based on multilayer networks, for the analysis of two rare disease scenarios with specific data granularities, effectively overcoming the classical constraints hindering rare disease and precision oncology research.
The first research article presents a personalized medicine study of the molecular determinants of severity in congenital myasthenic syndromes (CMS), a group of rare disorders of the neuromuscular junction (NMJ). The analysis of severity in rare diseases, despite its importance, is typically neglected due to data availability. In this study, modelling of biomedical knowledge via multilayer networks allowed understanding the functional implications of individual mutations in the cohort under study, as well as their relationships with the causal mutations of the disease and the different levels of severity observed. Moreover, the study presents experimental evidence of the role of a previously unsuspected gene in NMJ activity, validating the hypothetical role predicted using the newly introduced methodologies.
The second research article focuses on the applicability of multilayer networks for gene priorization. Enhancing concepts for the analysis of different data granularities firstly introduced in the previous article, the presented research provides a methodology based on the persistency of network community structures in a range of modularity resolution, effectively providing a new framework for gene priorization for patient stratification.
In summary, this PhD thesis presents major advances on the use of multilayer network-based approaches for the application of precision medicine to data-scarce scenarios, exploring the potential of integrating extensive available biomedical knowledge with patient-specific data.
ca
