Bioscience Method 2024, Vol.15, No.1, 37-49 http://bioscipublisher.com/index.php/bm 38 development trends of this process, it is hoped to provide useful reference and inspiration for researchers in the field of drug research and development, and promote the further development and application of AI based drug screening technology. 1 Overview of Drug Screening Process 1.1 Basic process and key steps of drug screening The drug screening process is the core link in drug development, which involves accurately selecting compounds with therapeutic potential from a massive selection of candidate drugs. This process typically begins with in-depth research on specific diseases or symptoms to clarify their biological mechanisms and potential drug targets (Figure 1). Researchers will screen potential active candidate drugs from a wide range of drug or compound libraries based on these targets. This preliminary screening process is usually completed through in vitro experiments such as high-throughput screening techniques and cell models to quickly evaluate the affinity and activity of candidate drugs with target molecules. Figure 1 Steps of drug screening Pan et al. (2019) developed a novel high-throughput screening method using gas chromatography-high-resolution mass spectrometry (GC-HRMS) technology for the screening of 288 drugs and toxins in human blood. This method allows for rapid detection and identification of many forensic important drugs and toxins, such as abused drugs (such as cocaine, amphetamines, synthetic cannabinoids, opioids, hallucinogens), sedatives and hypnotics, antidepressants, nonsteroidal anti-inflammatory drugs, insecticides (such as acaricides, fungicides, insecticides, nematicides), and cardiovascular drugs. Costa et al. (2019) developed a two-step drug screening process, including rapid screening by paper spray method, and then confirmed by liquid chromatography/mass spectrometry (LC/MS). This method demonstrates the potential application of testing drug compliance from fingerprints. This method first uses paper spray analysis to quickly screen a large number of samples, and then uses LC/MS to confirm any controversial results, especially the screening and confirmation of the antipsychotic drug quetopine, demonstrating the practicality of this method. Lin and Zhou (2022) proposed a drug candidate screening scheme based on machine learning methods to improve the efficiency of drug screening. This method can not only discover appropriate compounds, but also reveal the potential impact of molecular descriptors (i.e. feature values) on the properties of compounds. This work involves training an accurate prediction model based on independent variables (i.e. eigenvalues) and dependent variables (i.e. biological activity values or ADMET attributes), then using feature interpretation algorithms to select features that have a significant impact on dependent variables, finally finding approximate optimal values for these important features, and analyzing numerical ranges that are beneficial for obtaining better biological activity and ADMET attributes. However, relying solely on in vitro experiments is not sufficient. The in-depth screening stage requires researchers to further explore the activity, pharmacokinetic characteristics, and safety of candidate drugs in vivo. This stage of research is usually more complex and time-consuming, and requires the use of animal models or clinical trials to verify the actual effects of candidate drugs. During the entire screening process, selecting appropriate screening models, effectively analyzing and interpreting experimental data, and continuously optimizing the structure and properties of candidate drugs are all crucial steps. To ensure the accuracy and reliability of the screening results, researchers often need to conduct multiple rounds
RkJQdWJsaXNoZXIy MjQ4ODYzMg==