Machine Learning-Assisted Array-Based Biomolecular Sensing Using Surface-Functionalized Carbon Dots.

Area of Science:

• Analytical Chemistry
• Biotechnology
• Materials Science

Background:

• Fluorescent array-based sensing offers sensitive analyte detection in complex environments.
• Pattern recognition is key to differentiating analytes using optical patterns from sensor arrays.
• Machine learning (ML) presents a powerful, adaptive approach for analyzing complex sensing data.

Purpose of the Study:

• To develop an array-based sensor using carbon dots for differential protein detection.
• To demonstrate the efficacy of machine learning algorithms in analyzing fluorescence signals from the sensor array.
• To compare the performance of ML algorithms against classical statistical methods for pattern recognition.

Main Methods:

• Synthesis of carbon dots with varied surface functionality for sensor array fabrication.
• Utilizing fluorescent array-based sensing to generate optical patterns upon protein interaction.
• Applying machine learning algorithms, specifically Gradient-Boosted Trees, for pattern recognition and data analysis.
• Comparison with Linear Discriminant Analysis, a classical statistical method.

Main Results:

• The developed carbon dot sensor array successfully differentiated between eight different proteins at a 100 nM concentration.
• Machine learning algorithms, particularly Gradient-Boosted Trees, achieved 100% prediction efficiency in analyzing the sensing data.
• Gradient-Boosted Trees significantly outperformed the classical statistical method, Linear Discriminant Analysis.

Conclusions:

• Array-based sensing combined with machine learning offers a highly effective platform for sensitive and accurate analyte detection.
• Easy-to-synthesize carbon dots are suitable materials for developing versatile fluorescent sensor arrays.
• Machine learning provides superior capabilities for pattern recognition in complex fluorescent sensing data compared to traditional statistical methods.