Wired for Survival: Modeling Gene Networks to Identify Resilience and Neuron Subtypes in ALS via Graphical Optimization

Overview

Can neuron resilience be predicted as a topological failure of cellular identity? Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease marked by TDP-43 aggregation and motor neuron (MN) death. While current research emphasizes degeneration, mechanisms of neuronal resilience remain unclear. This study engineered a topologically-aware Graph Neural Network (GNN) to model ALS resilience through network stability. Using leave-one-out cross-validation, the model distinguished ALS from control with 85% accuracy. Applied to post-mortem MN samples, the GNN identified a 20-gene resilience hub enriched in chromatin remodeling and neuroprotection, with significantly higher activation in ALS (p=7.61e-04). A scRNA-seq dataset (GSE226482) was utilized in studying the epigenetic signature of the resilience hub. Seurat analysis revealed that MNs exist as subtypes along a state-specific spectrum; immature neurons highlight resilience hub activation, while mature, vulnerable MNs display end-stage disease markers. Type 7 was unique in possessing greater resilience even in a more mature state. These subtypes were validated in the large dataset: Type 7 signatures were enriched in long-surviving ALS samples (p=1.93e-03), and higher resilience hub expression correlated with increased survival (p=0.044). Mathematical network modeling using path-finding algorithms defined a “differentiation tax,” the cost of maintaining neuronal identity under stress, and identified efficient resilience networks with lower cost. By redefining ALS as topological failure, this research identifies a novel epigenetic shield for state-specific resilient neuronal subtypes. This graphical optimization framework implements a scalable method for identifying therapeutic targets that promote survival in motor neurons, providing hope for ALS patients.

Competition history

CSEF 2026 Medicine & Physiology (Track 2) (Senior Division) · Entry S-21-05

Related projects

ISEF · 2026