Sensing the Performance of Artificially Intelligent Nanopores Developed by Integrating Solid-State Nanopores with Machine Learning Methods

Abstract

Solid-state nanopores with a through-hole diameter of less than a few hundred nanometers can detect single nanoparticles owing to the ionic current flowing through the nanopore. Improvements in our understanding of the electrical properties of nanopores and the flow dynamics of a single nanoparticle passing through a nanopore have facilitated the use of machine learning methods for single nanoparticle detection. However, the sensing performance of solid-state nanopores integrated with machine learning methods has not been investigated to date. In this work, we reveal the sensing performance of artificially intelligent nanopores (AINs) comprising solid-state nanopores combined with a machine learning approach. An AIN with a diameter of 295 nm provided single nanoparticle identification with an accuracy of >91% based on single ionic current−time waveforms of 2−7 types of polystyrene nanoparticles with nominal diameters ranging from 90 to 300 nm. A 98% sample identification accuracy was achieved for the samples involving nanoparticles with diameters of 200 and 220 nm. Additionally, AINs can be used to develop multiplex diagnostics for infectious diseases owing to the capability of AINs to identify several viruses.

The New Era of High-Throughput Nanoelectrochemistry

“…Hu et al demonstrated the use of a machine learning algorithm to obtain automated classification of metabolites using a nanopore platform . Taniguchi et al reported the identification of several polystyrene nanoparticles by combining solid-state nanopore sensing with a machine learning algorithm . The same group has demonstrated that such approach can also be employed for accurate detection of different types of viruses .…”

“… 66 Taniguchi et al reported the identification of several polystyrene nanoparticles by combining solid-state nanopore sensing with a machine learning algorithm. 67 The same group has demonstrated that such approach can also be employed for accurate detection of different types of viruses. 68 A further example of a machine learning approach for discriminating analytes in mixtures is the use of the learning time-series shapelets, as exemplified by Wei et al 69 Here, the maximum discriminative features corresponding to each analyte that form a short segment of time-series data are extracted as shapelet signals and use the shapelet-transformed representation in order to classify the analytes investigated.…”

The new era of high throughput nanoelectrochemistry

“…68 Taniguchi et al reported the identification of several polystyrene nanoparticles by combining solid-state nanopore sensing with a machine learning algorithm. 69 The same group has demonstrated that such an approach can also be employed for accurate detection of different types of viruses. 70 A further example of a machine learning approach for discriminating analytes in mixtures is the use of the learning time-series shapelets, as exemplified by Wei et al.…”

Development of an Artificially Intelligent Nanopore for High-Throughput DNA Sequencing with a Machine-Learning-Aided Quantum-Tunneling Approach

“…Note that these ML models are transferable, are easy to access, and can be assembled with a nanopore sequencing device. Taniguchi et al have experimentally reported the implementation of AI solid-state nanopores with ML assistance and checked their sensing performance with ionic current–time data of nanoparticles . To tailor a suitable regressor model for the nanopore device, we have checked the prediction capability of all four optimized XGBR models, XGBR_dAMP , XGBR_dTMP , XGBR_dGMP , and XGBR_dCMP .…”