Deciphering DNA nucleotide sequences and their rotation dynamics with interpretable machine learning integrated C₃N...

Area of Science:

• Nanotechnology
• Quantum Physics
• Bioinformatics

Background:

• Solid-state nanopores with quantum transport show promise for rapid DNA sequencing.
• Precise single nucleotide analysis is challenging due to complex experimental protocols.

Purpose of the Study:

• To develop a machine learning (ML) framework to enhance single nucleotide recognition in DNA sequencing.
• To accelerate high-throughput analysis using C3N nanopores and quantum transport.

Main Methods:

• Utilized an optimized eXtreme Gradient Boosting Regression (XGBR) algorithm for nucleotide fingerprint transmission prediction.
• Employed SHapley Additive exPlanation (SHAP) for ML model interpretability.
• Performed comprehensive ML classification of nucleotides using binary, ternary, and quaternary combinations.

Main Results:

• XGBR achieved low root mean square error scores (as low as 0.07) for predicting nucleotide transmission and rotation dynamics.
• SHAP provided insights into ML model mechanisms and electrode-nucleotide coupling.
• ML classification reached maximum accuracy and F1 scores of 100%.

Conclusions:

• ML combined with nanopore devices can overcome quantum tunneling experimental hurdles.
• This integrated approach facilitates fast and high-precision DNA sequencing.
• Potential applications include disease diagnosis and personalized medicine.