Learning biomolecular absorption spectra in graphene nanopores

Area of Science:

• Biophysics
• Computational Chemistry
• Materials Science

Background:

• Solid-state nanopores offer potential for biomolecule detection when combined with optical measurements.
• Current methods for optical biosensing often require computationally intensive simulations or costly experiments.
• Developing reliable optical biosensing libraries is crucial for accurate biomolecule identification.

Purpose of the Study:

• To develop a computationally efficient learning approach for optical biosensing.
• To train deep neural networks on physical characteristics of amino acids within a graphene nanopore.
• To predict orientation-dependent absorption spectra for accurate biomolecule identification.

Main Methods:

• Utilized density functional theory (DFT) simulations to generate electronic, conformational, and optical data for single amino acids.
• Employed deep neural networks trained on DFT simulation data.
• Applied Principal Component Analysis (PCA) to compress high-dimensional data and improve training efficiency.

Main Results:

• The learning model accurately predicted orientation-dependent absorption spectra for five amino acids.
• Achieved high accuracy (test R^2 > 0.9) for previously unseen molecular orientations.
• Demonstrated robust interpolation across conformational variations within amino acid datasets.

Conclusions:

• The developed workflow enables the creation of reliable optical biosensing libraries with near-DFT accuracy.
• This approach significantly reduces computational cost compared to traditional simulations or experiments.
• Provides a scalable and efficient pathway for real-time, high-precision optical identification of biomolecules.