A Data-Driven and Interpretable Framework for Discovering New Analytical Detection Mechanisms in Surface-Enhanced...

Area of Science:

• Spectroscopy
• Chemical Analysis
• Machine Learning

Background:

• Surface-enhanced Raman spectroscopy (SERS) provides molecular fingerprints for chemical analysis.
• Quantitative SERS sensor development is often empirical, limiting efficiency and potentially missing key spectral features.

Purpose of the Study:

• To present a data-driven and interpretable framework (DIMS) for discovering SERS sensing indicators.
• To enable efficient identification of spectral features governing analyte quantification without prior manual selection.

Main Methods:

• High-throughput Raman spectral acquisition across varying analyte concentrations.
• Machine learning for spectral classification.
• Interpretable analysis to identify key spectral features for quantification.

Main Results:

• The DIMS framework successfully identified new spectral indicators for urea detection in artificial sweat.
• Enabled reliable qualitative and quantitative SERS sensing down to 10^-8 M.
• Demonstrated acceleration in discovering chemically meaningful sensing mechanisms.

Conclusions:

• DIMS offers an efficient strategy for developing high-performance SERS sensors.
• The methodology facilitates interpretable feature discovery and analysis of molecular interactions.
• The approach is extendable to other spectroscopic techniques.