National and Kapodistrian University of Athens
Discovery of new pharmaceutical approaches using protein interaction networks, molecular modelling and organic spectroscopy
Abstract
dc:descriptionThis thesis leverages advanced computational methodologies to predict and characterize the host–gut microbiome interplay through a structural lens, thereby addressing the inherent limitations of experimental interactomics and the scarcity of high-resolution structural data. The research is organized around three primary pillars: the development of predictive models, the biological characterization of the predicted interactome, and the translation of these findings into therapeutic insights. The first phase of this work focuses on deciphering the molecular interplay between the microbial and host proteomes. Initial efforts involved the construction of a protein–protein interaction network (PPIN) using machine learning (ML) to extract statistically significant associations between Pfam domains. This model was trained on a large-scale, heterogeneous dataset comprising over one million experimentally validated interactions across multiple taxa. To enhance predictive fidelity, a novel hybrid deep learning (DL) architecture was subsequently developed. This framework integrates graph-based protein representations with Variational Autoencoders (VAEs) to generate latent structural embeddings, which are then processed by a bi-directional cross-attention module to predict interactions. To ensure robustness against class imbalance, the model employs a focal loss that prioritizes hard-to-classify samples. The second objective involves the systematic biological and topological analysis of the predicted PPIN in order to identify disease associations. Through multi-parametric topological evaluations—including degree distributions, assortativity, and mean degree–abundance correlations—the study characterizes the resilience and organizational logic of the host–microbe interactome. This analysis identified critical protein hubs implicated in inflammation, cancer, mitochondrial dysfunction, and cardiometabolic disorders. The final phase translates these computational findings into pharmacological insights. Focusing on the identified hubs, the study employed an orthological design strategy to develop small organic molecules targeting pathological interactions in obesity- and cardiometabolic-related pathways. Given the strong association uncovered between the interactome and colorectal cancer (CRC), the DL model was further applied to clinical metagenomic data to distinguish the molecular signatures of healthy individuals from those of CRC patients. Centrality analysis identified key host proteins, which were then targeted for the rational repurposing of natural products. Achillea millefolium was identified as a primary candidate, and molecular docking together with nuclear magnetic resonance (NMR) spectroscopy were subsequently employed to validate the binding affinity of its synthetic compounds against the identified host targets. Overall, this thesis integrates computational modeling, structural biology, and clinical systems data to advance the understanding of the gut microbiome's influence on human health. By establishing a multi-scale pipeline—spanning from deep learning–based prediction to experimental validation—this work provides a foundational framework for discovering therapeutic targets and for elucidating the molecular mechanisms that underlie host–microbe cross-talk in both physiological and pathological states.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kiouri, Despoina
- Κιούρη, Δέσποινα
Subjects
dc:subject × 22- Μικροβίωμα του εντέρου
- Δίκτυα αλληλεπιδράσεων πρωτεϊνών
- Μηχανική μάθηση
- Μοριακή πρόσδεση
- Φασματοσκοπία πυρηνικού μαγνητικού συντονισμού
- Gut microbiome
- Protein–Protein Interaction Networks
- Machine learning
- Molecular docking
- Nuclear magnetic resonance spectroscopy
- Χημεία
- Φυσικές Επιστήμες
- Οργανική χημεία
- Βιολογία
- Μαθηματική και Υπολογιστική βιολογία
- Βιοτεχνολογία
- Chemical Sciences
- Natural Sciences
- Organic Chemistry
- Biological Sciences
- Mathematical and Computational Biology
- Biotechnology
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62653