Decoupling neurodegenerative protein signals in biofluids via cross-reactive interfaces and inverted multi-dynamic...

The direct analysis of neurodegenerative proteins in complex biofluids is inherently limited by severe Debye screening and non-specific protein interference, often necessitating extensive sample dilution in conventional assays. Traditional target-specific sensors fail to extract reliable signals from these physiological matrices. Here, we present a hardware-software co-design integrating a cross-reactive tiered deoxygenation graphene oxide (TDGO) sensor with an inverted multi-dynamic encoder (IMDE) framework to simultaneously quantify multiplexed targets (Aβ₄₂, tau-441, and α-synuclein). To overcome matrix complexity, the physical interface operates under a dual-stress condition, combining inherent matrix interference with dynamic electrochemical modulation (stepped back-gate voltage, V_BG = 0-20 V). This approach generates high-dimensional orthogonal data from non-linear sensor outputs, achieving a physical limit of detection below 0.1 pg/mL. By dynamically mapping these physical variations, the IMDE mathematically inverts the signal entanglement to achieve precise target quantification (R² = 0.914) directly in protein-fortified human serum. As a proof-of-concept for clinical application, we couple the IMDE with a multi-layer perceptron downstream classifier, successfully discriminating multi-class disease states in neat transgenic mouse plasma (AUROC = 0.917). Complemented by SHAP analysis to ensure signal interpretability, this platform provides a highly transferable bioelectronic methodology for decoding non-linear physiological signals.