Simultaneous Recognition and Detection of Adenosine Phosphates by Machine Learning Analysis for Surface-Enhanced...

Area of Science:

• Biochemistry
• Analytical Chemistry
• Spectroscopy

Background:

• Adenosine phosphates (AMP, ADP, ATP) are crucial for biological energy storage and signal transduction.
• Simultaneous detection of adenosine phosphates is vital for understanding cellular energy dynamics.
• Existing methods face challenges due to the structural similarities of adenosine phosphates.

Purpose of the Study:

• To develop a novel measurement system for simultaneous recognition and detection of adenosine phosphates.
• To overcome the limitations of current techniques requiring extensive training data.
• To enhance the accuracy of adenosine phosphate detection using limited spectral data.

Main Methods:

• Utilized surface-enhanced Raman scattering (SERS) with fabricated gold nanostructures for spectral measurements.
• Employed feature selection and data augmentation for preprocessing limited adenosine phosphate spectral data.
• Trained and compared various machine learning models, including multilayer perceptron.

Main Results:

• A multilayer perceptron model was successfully trained on augmented spectral data.
• The model achieved a high accuracy of 0.914 in detecting AMP, ADP, and ATP.
• Demonstrated effective recognition and detection of adenosine phosphates from limited SERS data.

Conclusions:

• Established a novel measurement system for accurate adenosine phosphate detection.
• Showcased the potential of combining SERS with machine learning for complex molecular analysis.
• Overcame the challenge of limited training data in spectroscopic analysis.