Computational Molecular Biology 2024, Vol.14, No.1, 9-19 http://bioscipublisher.com/index.php/cmb 11 In the molecular screening and optimization stage, artificial intelligence technology helps researchers quickly select potentially active candidate molecules from a huge compound library by building prediction models, and optimizes the structure of these molecules to improve their efficacy and reduce side effects. For example, the virtual screening method based on machine learning (Figure 1) can use the chemical structure and biological activity data of known active molecules to build a prediction model to predict and rank the activity of new molecules, thereby quickly screening out candidate molecules with potential activity. Figure 1 Virtual screening (Zhang et al., 2024) Artificial intelligence technology also plays an important role in the clinical trial stage. For example, DeepMind collaborated with Moorfields Eye Hospital to develop the Streams system, which uses deep learning technology to analyze eye scan images, automatically identify and interpret complex images, and provide preliminary diagnostic recommendations (https://zhuanlan.zhihu.com/p/41970785). This technology helps solve the problem of scarce expert resources and allows patients to receive timely diagnosis. In addition, prediction models based on machine learning can also predict and evaluate the results of clinical trials, providing strong support for the design and optimization of clinical trials. These applications can not only improve the efficiency and accuracy of clinical trials, but also help reduce the costs and risks of clinical trials. Zhang et al. (2022) found that despite a large investment of money and time, the success rate of clinical testing is still less than 15%. Approximately 50% of drug discovery failures are due to poor pharmacokinetic properties (absorption, distribution, metabolism, excretion and toxicity). With the development of computational methods, the speed and success rate of drug discovery have greatly improved.
RkJQdWJsaXNoZXIy MjQ4ODYzNA==