The AI Revolution: How is AI reshaping new pesticide research and development

The No.1 Central Document in 2024 puts the production of grain and important agricultural products in the first place. In today's agricultural production, the application of pesticides is one of the key measures to improve crop yield and ensure food security, but the "toxic leek" and other incidents reflect the safety of pesticides, as well as the irrational use of pesticides caused by resistance and other problems are always difficult to be solved. With the rapid development of artificial intelligence (AI), the combination of AI and many original industries has sprung up.

Use artificial intelligence to identify pesticide risks in advance 

The "toxic leek" incident in 2010 was essentially organophosphorus poisoning caused by eating leek with excessive pesticide residues. Although the current standards for pesticide residues in our country are increasingly strict, the irrational use of pesticides still cannot avoid the occurrence of such problems.

From the perspective of pesticide development process, pesticide toxicological evaluation is usually carried out before submission for pesticide registration, which results in many candidate compounds with excellent activity being found to have safety problems before nearby registration, which ultimately leads to registration failure. This undoubtedly cost a lot of research and development money and time.

Artificial intelligence can conduct modeling analysis by supervising learning from existing toxicity data of compounds. This enables predictive analysis of unknown toxic compounds at the early stage of development. At present, there are several pesticide safety prediction models for pesticide genotoxicity, acute toxicity of fish, acute oral toxicity of rats, acute oral toxicity of mice, acute percutaneous toxicity of rats, acute percutaneous toxicity of rabbits, acute inhalation of mice, acute inhalation of rats, and acute contact toxicity of bees. This can provide early warning for pesticide developers of structural safety risks of compounds, and avoid the occurrence of similar events such as "toxic leek" from the root.

Artificial intelligence helps discover novel structural pesticides 

As long as pesticides continue to be used, the problem of resistance is inevitable. At present, the solution to the problem of resistance is usually how to use pesticides rationally, such as: rotation of pesticides, mixed pesticides and so on.

From the perspective of pesticide research and development, even if pesticides with similar structure types are successfully registered and listed, they may soon have extremely high resistance problems, such as Chlorantraniliprole, a pesticide developed by DuPont for diamond-moth control, which has been on the market for only 3 years. This means that novel structural types of pesticides need to be found to circumvent potential resistance risks.

The existing compounds that can be synthesized are of the order of 1060, and it is obviously unrealistic to find highly active compounds by experimental methods. Therefore, novel structure types can be found quickly by clustering the structure of compound libraries. At the same time, combined with quantitative pesticide model, it helps researchers to further select lead compounds with more potential pesticide for further development and research.

At present, Professor Yang Guangfu's team of Central China Normal University has combined machine learning methods such as QED, RDL and deep learning methods to establish COPLE artificial intelligence pesticide property prediction platform and Professor Zhang Li's team of China Agricultural University established agricultural activity prediction platform based on RF and SVM. The above prediction platforms can be used free of charge. This greatly improves the structural novelty rate of pesticide lead compounds, and to a certain extent reduces the risk of rapid resistance after the registration and listing of new pesticides.

A successful pesticide not only needs to have high activity against diseases, pests and grasses, but also needs to be safe and non-toxic to the environment, mammals and non-target organisms, and also needs to have low production costs and resistance risks. Artificial intelligence technology can, to a certain extent, help pesticide developers predict unqualified candidate compounds in the early stage of research and development, and reduce the safety risks in the later stage of new pesticide registration and the risk of resistance during the field use of pesticides. In short, the application of AI in the field of pesticides will greatly accelerate the research and development of green pesticides and promote the development of agricultural chemistry 4.0.