State Institution "Kundiiev Institute of Occupational Health of the National Academy of Medical Sciences of Ukraine", Kyiv, Ukraine
Introduction.The work discusses the use of pesticides in the agriculture of Ukraine. It was stated that in 2020, 46.2 million hectares used 40.7 thousand tons of pesticides, among which more than 2500 drugs were registered, including about 250 active substances. The growth of the use of complex drugs with several active substances and an increase in the load of pesticides on the hectare compared to 1997 is emphasized. Particular attention is paid to the need to develop methods for determining the remains of pesticides in products, feeds and the environment. Classical analytical methods are discussed, such as spectrophotometry, gas and liquid chromatography, as well as their restrictions related to the identification of new substances and mixed drugs. The importance of new mathematical honey processing of primary signals is noted to increase the accuracy of the quantitative analysis.
Objective. The Laboratory of Analytical Chemistry and Toxic Substances Monitoring aims to develop scientific, innovative ways to improve existing analytical chemistry methods for the control of industrial environments and environmental pollutants, in particular pesticides, taking into account the global spread of persistent organic pollutants (POPs), including organochlorines.
Materials and methods. Approaches to solving the set tasks are based on the application of machine intelligence methods to the processing of experimental materials obtained by gas-liquid and high-performance liquid chromatography, thin-layer chromatography (TLC), and spectrophotometry. Modern computing platforms were used for mathematical modeling of machine methods for recognition, classification and identification of signal-descriptors of molecules of investigated substances: Matlab 2019b, Gaussian 9.0, R V 4.1.3.
The results.The laboratory has been updating reference databases, like NIST, with mass spectra of modern pesticides. It has developed mathematical methods, such as OLAM-Decomposition, for separating overlapping chromatographic peaks of isomers and other compounds. Modern machine learning algorithms have been employed to classify chemical compounds based on their mass spectra, enabling the identification of substances not listed in existing databases. Techniques like chromatography-mass spectrometry combined with infrared spectrophotometry have increased the reliability of compound identification by comparing synthesized quantum mechanical models with experimental spectra. In biomonitoring, the laboratory developed a polynomial neural network-based model to predict the transfer of HCH isomers from mothers to prenatal children. For improved quantitative TLC analysis of low-quality chromatograms, machine learning methods enhanced accuracy, reducing error rates significantly. Research has expanded to include electrochemical impedance spectroscopy (EIS) for detecting pollutants like organochlorine pesticides and heavy metals in water. The laboratory collaborates with leading Ukrainian universities, involving students in innovative research, including automated molecule classifiers and EIS applications. Several student works have been recognized in national competitions.
Keywords: gas-liquid chromatography, high-performance liquid chromatography, thin layer chromatography, electrochemical impedance spectroscopy, polynomial neural networks, deep learning
References