Machine Learning Analysis for Phenolic Compound Monitoring Using a Mobile Phone-Based ECL Sensor.

Area of Science:

• Analytical Chemistry
• Sensor Technology
• Computational Science

Background:

• Low-cost sensors for point-of-care diagnostics face challenges like non-linearity and sensor variability.
• Machine learning (ML) offers a powerful approach to address these sensor data complexities.

Purpose of the Study:

• To develop and validate ML models for detecting and quantifying phenolic compounds.
• To investigate the modeling of electrochemiluminescence (ECL) quenching mechanisms using ML algorithms.
• To assess the performance of ML on mobile phone-based ECL sensor data.

Main Methods:

• Utilized ML algorithms including neural networks and Gaussian Process Regression.
• Modeled the ECL quenching mechanism of the [Ru(bpy)3]2+/TPrA system.
• Analyzed time series data and extracted features from a mobile phone-based ECL sensor.

Main Results:

• ML regression using minimally processed time series data showed high detection performance.
• Combined multimodal sensor data with multilayer neural networks improved performance by 80% compared to single-feature analysis.
• Demonstrated robust analysis of noisy and variable sensor data.

Conclusions:

• ML provides a robust framework for analyzing complex sensor data, overcoming limitations like noise and variability.
• ML strategies are crucial for chemical and biosensor data analysis, integrating nonlinearity and sensor variations for improved performance.