Inferencia de módulos transcripcionales desregulados específicos al paciente mediante redes neuronales gráficas.

Garrido Ochoa, David
Veleiro Carril, Uxía (dir.) ; Hernáez Arrazola, Mikel (dir.)

Universidad de Zaragoza, CIEN, 2024
Departamento de Bioquímica y Biología Molecular y Celular,

Máster Universitario en Biofísica y Biotecnología Cuantitativa / Master in Biophysics and Quantitative Biotechnology

Resumen: The main objective of personalized medicine is to adapt medical treatments to the individual profile of each patient. It relies on the identification of specific molecular signatures that drive disease progression. Traditional methods for identifying these signatures are based on statistical associations and lack interpretability, limiting their applicability. In this work, we propose a state-of-the-art Graph Neural Network (GNN) approach to infer patient-specific, dysregulated transcriptional modules for breast cancer patients.
GNNs provide a robust approach for modeling the complex relationships between genes, characterized by the interactions between the proteins they encode. Our methodology consists of two key components: i) a GNN model that learns to predict the metastatic phenotype from the gene expression data of a patient, and ii) an explainability module that identifies the most informative genes and interactions that drive the model's decision. Using publicly available breast cancer data, we demonstrate the effectiveness of our approach in identifying deregulated modules associated with clinical outcomes and how these findings can guide therapeutic decisions.
This thesis contributes to the advancement of personalized medicine by introducing a framework to discover patient-specific molecular signatures. Our procedure has the potential to enhance the diagnosis, treatment, and prognosis of diseases by providing insights into personalized therapeutic targets.

Tipo de Trabajo Académico: Trabajo Fin de Master