Detection for viruses before and during a pandemic

A September 2008 paper in the IEEE Sensors Journal, outlined sensor applications for early warning systems for bioterrorism. This effort was targeted at viruses with the ability of rapidly spreading in humans during a pandemic. Variants of such pathogens are difficult to detect, so a reliable detection strategy that could classify viruses from innocuous, harmless proteins is demonstrated in this article.

I am not an expert in this field, but I have to believe that this kind of detection process must have some use in our current COVID-19 pandemic. For all I know, maybe a similar technology is already being used or considered. Anyway, I thought the following explanation might be of great interest to all people interested in ideas/techniques to rid the world of this coronavirus

Instrumentation was developed based upon pyrolysis gas chromatography differential mobility spectrometry (PY/GC/DMS). This technique, coupled with a genetic algorithm that is based upon a machine learning (ML) technique for pattern recognition with a micro-fabricated solid state biosensor will be shown below.

Figure 1 The block diagram indicates a virus/biological sample entering on the left side of this block diagram and the Nitrogen N2 flow according to the arrow directions and amounts in the image. The output on the right shows that both positive and negative chemical spectral signals have been acquired (Image from Reference 1)

The study showed that this type of portable sensor is able to be used coupled with the sophisticated data mining method as a quick, reliable, and accurate tool for the recognition of viruses.

Our interest here will be confined to the sensor and the electronics. Reference 1 can be accessed on the IEEE Xplore site behind a pay wall.

Pyrolysis breaks down large complex hydrocarbon molecules of biomass into smaller/simpler molecules of gas, liquid, and char (a darkened crust). Figure 2.

Figure 2 A multi-functional pyrolysis system diagram (Image from Frontier Lab)

Differential mobility spectrometry is a technology for analyzing challenging biological samples via detection and characterization of gas-phase ions. Figure 3.

Figure 3 A high amplitude, high frequency, asymmetric waveform (known as separation voltage, SV) is applied to the electrodes to create an electric field (known as separation field, S) affecting the ions of the sample. (Image from Reference 2)

Machine Learning (ML)

Using ML, the data analysis performed calculations in three areas:

• Genetic Algorithm (GA) calculated and selected the most perceptible biomarkers
• Principle component analysis (PCA) transferred the original space, which spanned by the selected biomarkers, to a transitional space spanned by a number of principal components of the biomarker
• Backpropagation neural networks (BPNN) created a classifier that would predict the class index of each sample for the principal components in their biomarker

These three sections were integrated into a ML system which was dominated by the GA, which mimics a natural evolution principle. Figure 2 shows the flowchart of this ML process. Figure 4.

Figure 4 The flowchart of the genetic algorithm-based biomarker selection

This architecture has the ability to enable a high accuracy separation of the biological samples, and the sensor platform, coupled with the ML strategy may be widely applied to a broad spectrum of virus detection problems for both public health professionals and clinicians.

Written for @SupplyFrame

Reference

1 Differentiation of Proteins and Viruses Using Pyrolysis Gas Chromatography Differential Mobility Spectrometry (PY/GC/DMS) and Pattern Recognition, Susan Ayer, Weixiang Zhao, and Cristina E. Davis, IEEE SENSORS JOURNAL, VOL. 8, NO. 9, SEPTEMBER 2008

2 Differential mobility spectrometer: Model of operation, E Krylov, E G Nazarov, R A Miller, International Journal of Mass Spectrometry 266(1):76–85, September 2007