Phd in HUMAN BIOLOGY AND MEDICAL GENETICS

Thesis title: Dynamic Network Signatures from Molecular Dynamic Simulation: A Framework for Rare Genetic Diseases

The use of next-generation sequencing (NGS) technologies has significantly increased the number of genetic variants identified in diagnostics and research. This has led to an exponential rise in variants of uncertain significance (VUS), creating a bottleneck in the molecular diagnosis of rare genetic diseases and beyond. While Molecular Dynamics (MD) can theoretically reveal how these variants reshape protein behavior to aid interpretation, omparing large ensembles of simulations in a consistent way remains a challenge. Addressing the need to bridge the gap between static genotype and dynamic phenotype, this thesis introduces NetMD, an unsupervised, graph-based Artificial Intelligence (AI) framework. NetMD converts MD trajectories into evolving “network signatures” of residue interactions and aligns them to uncover shared pathways and variant-specific departures without requiring labels or predefined reaction coordinates. While motivated by the challenge of VUS interpretation, we validated the framework by applying it to characterize known pathogenic variants and pharmacological mechanisms across diverse biological contexts. In GLUT1 (implicated in GLUT1 Deficiency Syndrome and cancer metabolism), NetMD differentiated wild-type from pathogenic variants (R333Q/W), highlighting shifts toward inactive-like transport states. Extending the analysis to oncology, we aligned inhibitor-binding trajectories for reference GLUT1 blockers, capturing binding-induced network changes consistent with the stabilization of inward-open ensembles. In KDM6A (associated with Kabuki syndrome), the method revealed how the pathogenic R1255W variant perturbs long-range couplings across the catalytic domain. Finally, in mitochondrial Complex I, focusing on the ND6 M64V variant causing Leber Hereditary Optic Neuropathy (LHON), NetMD exposed the rewiring of conserved subunit interfaces. By summarizing complex simulations into clear, time-resolved network patterns, NetMD provides a robust tool to define the dynamic signatures of pathogenicity and drug response, offering a scalable route to support variant classification and rational therapeutic targeting.