Machine learning-assisted dual-mode intelligent biosensor for sarcosine detection based on fluorescent nanozymes

The development of sensitive, cost-effective, and reliable biosensors for early prostate cancer screening remains a pressing challenge. Here, we report a machine learning-assisted dual-mode biosensing platform based on iron-doped carbon quantum dots (Fe_0.45-CDs) coupled with sarcosine oxidase (SOX) for the ultrasensitive detection of sarcosine (Sar), an emerging biomarker for prostate cancer. The Fe_0.45-CDs synthesized via a rapid microwave-assisted strategy possess unique structural and optical advantages, including abundant surface functional groups, strong blue fluorescence, and robust peroxidase-like catalytic activity. These merits facilitate the construction of a cascade reaction-based sensor integrating colorimetric and ratiometric fluorescent readouts, which mitigates environmental interference and improves detection sensitivity. The proposed dual-mode platform delivers limits of detection 0.922 μM (colorimetry) and 2.34 μM (fluorometry), outperforming conventional single-mode nanozyme biosensors. Furthermore, RGB features extracted from smartphone-captured images were analyzed using multiple machine learning algorithms, among which the k-nearest neighbors (KNN) model achieves optimal prediction accuracy toward Sar concentration. Importantly, favorable anti-interference capability and satisfactory spike recoveries are verified in urine samples, demonstrating promising clinical practicability. This work demonstrates the rational design of doped carbon dot nanozymes as versatile sensing elements and validates the synergistic combination of nanozyme-bioenzyme cascade catalysis and artificial intelligence to advance noninvasive cancer diagnosis and biomarker quantification.