2023-09-01 06:30:38
What biotech innovation should we not miss this month? A pesticide screening method made possible by artificial intelligence…
Excess pesticides in food are a serious danger to public health. A rapid and upstream detection of possible residues is therefore crucial. Especially when it comes to foods from mass distribution such as vegetables, with a limited shelf life. Yuanhao Zhou, from Sun Yat-sen University in Shenzen (China), and his Chinese colleagues have therefore worked on a new method for detecting pesticides on the surface of plants. For this, they used a miniature mass spectrometer, capable of detecting and identifying the molecules of interest by simply measuring their mass. They then coupled it with a deep learning algorithm that they developed with the aim of sorting the results efficiently and autonomously. The entire process is presented in detail in the issue of Food and Chemical Toxicology dated August 28, 2023.
Pesticides spotted “every time”
Before being analyzed by the mass spectrometer, the vegetable to be screened must first be pretreated. It is actually a matter of bringing out the pesticides it contains to make them detectable by the machine. Chinese scientists carried out various tests involving five different pesticide residues – carbendazim (biocide), dimethomorph and azoxystrobin (fungicides), tebufenozide and cyromazine (insecticides) – spread on four types of plants (cowpea, leek, celery and pepper ). According to them, the optimal method of pre-treatment would be to place the vegetable in a flexible airtight bag. The best material for the latter would be polyamide mixed with cast polypropylene, as these components interfere little with the targeted compounds. The bag is also filled with a solution composed of methanol and water in a 1:1 ratio (its volume in mL must equal the mass of the plant in g). The role of the solution, or eluent, will be to extract the plant residues. After shaking everything for one minute, the liquid is filtered through a nylon membrane (0.22 μm mesh) before passing to spectrometry.
Previously, other detection methods had been used: nuclear magnetic resonance, Raman spectroscopy, immuno-enzymatic method… But spectrometry is the one that has shown the most promising results in the laboratory: high specificity and sensitivity, and precision both in quality than in quantity. Problem: Mass spectrometers are bulky and complicated to operate. This is why miniature versions were invented and are currently employed in environmental protection, public safety and even during medical examinations. However, miniaturization is accompanied by a noticeable drop in performance… Fortunately for Yuanhao Zhou and his team, the development of artificial intelligence has led to the development of machine learning algorithms applied to reduced-size mass spectrometers. Some have, for example, been used for fruit recognition or for the diagnosis of parasitosis in poultry. The Chinese researchers, on the other hand, designed their algorithm with a view to improving the spectrometer’s on-site detection capabilities. A successful bet since the latter was able to identify quantities of pesticides as small as 10 μg/kg of vegetable, with an accuracy of 99.62%!
1693562243
#Spotting #pesticides #deep #learning