Simulation-Driven Biomarker Selection for Graphene-Based Stroke Biosensors: A Multiscale Computational Framework

Early and accurate detection of acute stroke remains a major clinical challenge due to the narrow therapeutic window for effective intervention. This study presents a simulation-driven computational framework for biomarker selection and biosensor-oriented modeling, aimed at guiding the development of graphene-based diagnostic platforms. Two biomarker panels were evaluated, comprising inflammatory cytokines (IL-۶, TNF-α, and CRP) and neurological injury markers (GFAP, NSE, and MMP-۹). A synthetic dataset (N = ۵۰۰), generated from literature-informed statistical distributions and shown to be consistent with reported clinical trends, enabled physiologically consistent modeling of biomarker behavior and probabilistic classification using logistic regression. Interaction network analysis revealed stronger coupling among neurological biomarkers, reflecting coordinated pathological processes associated with early brain injury. Diagnostic evaluation demonstrated superior performance for the neurological panel, achieving an AUC of ۰.۹۱, sensitivity of ۰.۸۶, specificity of ۰.۸۳, and overall accuracy of ۰.۸۵, compared to the inflammatory panel (AUC = ۰.۸۲, accuracy = ۰.۷۶). These findings are consistent with previous studies highlighting the higher specificity of neurological biomarkers in early stroke detection. Importantly, the proposed framework explicitly bridges molecular-level interaction insights derived from prior molecular dynamics studies on graphene-based systems with system-level diagnostic modeling, enabling systematic simulation-driven pre-screening and prioritization of biomarker panels prior to experimental validation. This approach establishes a scalable, simulation-based paradigm that reduces experimental burden and supports the rational design of graphene-based biosensing platforms for early stroke detection.

کلیدواژه ها:

Stroke biomarkers ، Simulation-driven pre-screening ، Graphene-based biosensing ، Computational Modeling ، Probabilistic classification ، Diagnostic performance