Machine-Learning-Assisted Triple-Gated Raman Enhancement Platform for Selectively Quantifying Lysophosphatidylcholine (16:0) as a Potential Biomarker for Cognitive Impairments.

Huang, Xinwei; Miao, Qingyuan; Li, Yawei; Cong, Peilin; Feng, Ban; Li, Cheng; Dan, Tingting; Li, Xinyang et al. · ACS Nano · 2026

basic_science · Level V

Where this comes from

Abstract

Accurate quantification of structurally similar metabolites as biomarkers in biofluids has remained a longstanding challenge. Here, we report a semiconductor-organic hybrid interface (ZrS<sub>2</sub>@ZrO<sub><i>x</i></sub>-C16) with a triple-gated molecular recognition environment for high-specificity detection of lysophosphatidylcholine (16:0) (LysoPC (16:0)), which is identified as a potential biomarker associated with aging and cognitive decline. Through integrating phosphocholine-selective Zr-O-P coordination, chain-length-matched hydrophobic free-energy minimization, and a dual-resonant charge-transfer pathway, ZrS<sub>2</sub>@ZrO<sub><i>x</i></sub>-C16 affords molecular-level discrimination among lysophospholipids with nearly identical chemical structures, enabling amplified and selective quantitative Raman signals. Coupled with machine-learning extraction of Raman fingerprints, ZrS<sub>2</sub>@ZrO<sub><i>x</i></sub>-C16 achieves rapid, label-free quantification with an accuracy of <i>R</i><sup>2</sup> = 0.999 across human and mouse serum samples, allowing precise mapping of LysoPC (16:0) deficits as a biomarker and therapeutic target across aging, Alzheimer's disease, and perioperative neurocognitive impairment. This work establishes a framework for precision lipid analytics and high-selectivity metabolic sensing, enabling mechanistic insights in neurometabolic biology.