Smart sensing of creatinine in urine samples: Leveraging Cu-nanowires/MoS2 quantum dots and machine learning

Abstract

Serum creatinine (CRT) levels are key biomarkers for diagnosing, staging, and monitoring renal disease in clinical practice. In this work, copper nanowires (CuNW), and Molybdenum disulfide quantum dots (MSQD) modified glassy carbon electrode (GCE) were chosen to demonstrate the electrochemical detection of CRT in complex mixture and urine samples. The materials were characterized using various physical characterizations such as FESEM, XRD, UV, PL, and FT-Raman. The electrocatalytic activity of the sensor was investigated using cyclic voltammetry (CV), and differential pulse voltammetry (DPVs). Despite the elevated sensitivity and cost-effectiveness of electrochemical sensors, the performance of the sensors is constrained by the existence of interfering species that generate conflicting and overlapping electrochemical signatures. In order to address this issue, we implemented a machine learning (ML) approach to accurately quantify CRT levels in complex mixtures, as well as in urine samples. The ML algorithms employed are trained and tested on a large dataset, allowing them to effectively capture and analyze the variance in the electrochemical signatures, demonstrating the application of artificial intelligence. The proposed sensor exhibits linearity from 1.96 μM to 966.0 μM and shows the best performance in terms of limit-of-detection (LOD) of 2.3 μM in a complex mixture and 0.001 μM in real urine samples, with RMSE of 0.2 and 0.017 μM using artificial neural network and random forest ML models respectively. We anticipate that by further miniaturization of these sensors into point-of-care testing devices, renal diseases can be managed effectively.

Keywords

• Electrochemical sensor
• Copper nanowires
• Artificial neural networks
• Machine learning
• Creatinine