Retinal diseases (RD), which often cause irreversible vision impairment and lack effective treatments, necessitate more attention than common ocular conditions like cataract (CA), and aqueous humor metabolic analysis has proven an invaluable tool for their investigation. In this work, a robust, high-performance heterojunction nanoconstruct (MZ-D) with cost-effectiveness (∼2.7 yuan per gram), streamlined two-step synthesis, high reproducibility, was developed. Combined with high-throughput mass spectrometry and machine learning, aqueous humor metabolic profiles were extracted and analyzed for multiple RD, including age-related macular degeneration (AMD), diabetic retinopathy (DR), and retinal vein occlusion (RVO). We identified key metabolic features that delivered 92.87% accuracy, 0.9340 AUC, and 92.94% precision for RD recognition against CA, which was used as a control to simulate clinical settings, and 86.59% accuracy, 0.9591 AUC, and 86.59% precision for differentiating AMD, DR, and RVO. Notably, retinaldehyde and 2-hydroxylauroylcarnitine were unveiled as specific metabolic signatures of RD, and beyond playing a pivotal role in robustly distinguishing RD from CA, they exhibited significant discriminatory power across multiple RD subtypes, yielding accuracy and precision both exceeding 80%. This work paves the way for clinical translation of nanomaterial-enabled ocular metabolic diagnostics.
Lin et al. (Tue,) studied this question.