Deep Learning-Enhanced Nanopore Sensing of Single-Nanoparticle Translocation Dynamics

Area of Science:

• Nanotechnology
• Machine Learning
• Analytical Chemistry

Background:

• Electrical sensors often struggle with ubiquitous noise, hindering the detection of faint but crucial signals.
• Resistive pulse sensing is a technique used to detect nanoparticles, but it is susceptible to noise interference.

Purpose of the Study:

• To develop and demonstrate a deep learning approach for denoising ionic current signals in resistive pulse sensing.
• To enable the detection of subtle nanopore corrugation signals and track fast-moving nanoparticles.

Main Methods:

• A convolutional auto-encoding neural network was designed for denoising ionic current waveforms.
• The network was trained using gradient descent optimization to minimize waveform differences.
• The method was applied to detect electrophoretically-driven single-nanoparticle translocation in a nano-corrugated nanopore.

Main Results:

• The deep learning model successfully reduced noise in ionic current data.
• Denoising enabled the identification of corrugation-derived wavy signals previously undetectable.
• In-situ tracking and electrokinetic analysis of fast-moving single- and double-nanoparticles became feasible.

Conclusions:

• Deep learning offers a powerful method for denoising in solid-state nanopore sensing.
• This approach preserves temporal resolution, crucial for analyzing dynamic biological molecules.
• The technique holds promise for applications in protein structure and polynucleotide sequence analysis.