利用多尺度知识图谱框架绘制ALS中趋同的神经退行性与致癌通路

Amyotrophic lateral sclerosis (ALS) affects approximately 1.5 per 100,000 person-years in the United States, with a point prevalence of 3.8-5.6 per 100,000. ALS arises from the intersection of diverse pathogenic mechanisms spanning genetic mutations (C9orf72, SOD1, FUS, TARDBP), protein misfolding, RNA-metabolism defects, nucleocytoplasmic transport failure, prion-like propagation, mitochondrial dysfunction, and neuroinflammation. Increasing evidence from plasma proteomics and machine-learning studies indicates that these perturbations emerge years prior to clinical onset, highlighting an urgent need for mechanistically grounded patient stratification. At the same time, epidemiologic and molecular studies suggest unexpected parallels between ALS and cancer-including dysregulated DNA-damage responses, cell-cycle control, metabolic rewiring, and aberrant immune signaling-yet these cross-disease links remain fragmented and poorly mapped. We developed a multi-scale ALS-cancer knowledge graph integrating genetic, transcriptomic, proteomic, neuropathological, and longitudinal clinical datasets. Knowledge-graph completion approaches, including link prediction, graph embeddings, and embedding-based reasoning, were applied to infer missing edges and reveal latent modules connecting ALS-associated genetic lesions to oncogenic pathways implicated in genome instability, altered proliferation, metabolic remodeling, and immune dysregulation. This approach enables systematic interrogation of shared molecular programs across ALS and cancer, exposing under-characterized immune, metabolic, and axonal-transport circuits whose dual roles have been suggested but never comprehensively mapped. Predicted associations were evaluated using enrichment for established ALS genes, overlap with curated oncogenic and neurodegenerative pathways, and their ability to stratify patients by site of onset, progression rate, and biomarker signatures. These analyses generated patient-specific pathway fingerprints that (i) resolve dominant biological mechanisms across clinical subgroups, (ii) illuminate mechanistic intersections between neurodegeneration and malignancy, and (iii) prioritize repurposed oncology targets with potential relevance to ALS. This knowledge-graph framework provides a scalable and mechanistically informed strategy for classifying ALS patients, uncovering ALS-cancer pathway convergence, and supporting targeted therapeutic discovery across traditionally siloed disease domains.