Identifying Hub Genes of Alzheimer’s and Parkinson’s Diseases via h-cutoff: A New Methodological Approach

Read the full article

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Background: Alzheimer’s disease (AD) and Parkinson’s disease (PD) are genetically complex neurodegenerative disorders in which disease-associated genes act through coordinated immune, glial, proteostatic, metabolic, synaptic and vascular programs rather than through isolated single-gene effects. Weighted co-expression and co-occurrence networks are therefore attractive for prioritizing disease genes, but conventional hard thresholds can remove low-weight edges that carry high structural importance. Objective: This manuscript refines the original h-metrics terminology into an explicit h-cutoff framework for weighted biological networks. The revised procedure uses co-expression or co-occurrence values as edge weights, applies a one-order h-cutoff to obtain an h-subnet, adds weak bridge edges to form an h-backbone, and then applies a two-order h-cutoff within the h-backbone to identify final Hub genes. Methods: The method was formalized for non-integer edge weights by extending the integer h-strength definition to a continuous h-strength through scaling and interpolation in rank space. The analytical design was organized into AD/PD-shared, AD-specific and PD-specific layers using public resources, including GSE48350, GSE26927, GSE174367, GSE161045, NIAGADS and related AD/PD transcriptomic and genetic studies. The mathematical workflow was written as a reproducible edge-to-backbone-to-gene pipeline with numbered formulas, clearly separating functional edge strength, structural weak-bridge value and final node-level Hub gene selection. Results: The shared AD/PD h-backbone prioritized CXCR4, FLT1, HSPB1, HSPA1A, CALM3, CDC42, RAB3A, TREM2 and APOE, supporting a convergent neurodegenerative architecture involving neuroinflammation, proteostasis, glial lipid biology and synaptic/cytoskeletal remodeling. The AD-specific layer prioritized DLAT, CCDC88B, SREBF1, APOE, CLU, TREM2 and TYROBP, consistent with mitochondrial stress and glial lipid/immune regulation. The PD-specific layer prioritized PSMB8, GRIA1, RGS8, BAG3, SYN1, CALB2, SNCA and TH, consistent with proteasome/chaperone stress, synaptic signaling, alpha-synuclein biology and dopaminergic vulnerability. Conclusions: Continuous h-cutoff provides an adaptive, interpretable and reproducible alternative to arbitrary correlation thresholding. It preserves the original h-backbone principle of integrating functional and structural network information while adapting h-strength to non-integer biological edge weights. The revised method is suitable for gene-prioritization studies in AD, PD and other complex polygenic diseases.

Article activity feed