Machine learning powered detection of biological toxins in association with confined lateral flow immunoassay (c-LFA).
Journal:
The Analyst
PMID:
39101439
Abstract
Biological weapons, primarily dispersed as aerosols, can spread not only to the targeted area but also to adjacent regions following the movement of air driven by wind. Thus, there is a growing demand for toxin analysis because biological weapons are among the most influential and destructive. Specifically, such a technique should be hand-held, rapid, and easy to use because current methods require more time and well-trained personnel. Our study demonstrates the use of a novel lateral flow immunoassay, which has a confined structure like a double barbell in the detection area (so called c-LFA) for toxin detection such as staphylococcal enterotoxin B (SEB), ricinus communis (Ricin), and botulinum neurotoxin type A (BoNT-A). Additionally, we have explored the integration of machine learning (ML), specifically, a toxin chip boosting (TOCBoost) hybrid algorithm for improved sensitivity and specificity. Consequently, the ML powered c-LFA concurrently categorized three biological toxin types with an average accuracy as high as 95.5%. To our knowledge, the sensor proposed in this study is the first attempt to utilize ML for the assessment of toxins. The advent of the c-LFA orchestrated a paradigm shift by furnishing a versatile and robust platform for the rapid, on-site detection of various toxins, including SEB, Ricin, and BoNT-A. Our platform enables accessible and on-site toxin monitoring for non-experts and can potentially be applied to biosecurity.