Computational Molecular Biology 2024, Vol.14, No.3, 115-124 http://bioscipublisher.com/index.php/cmb 119 5.1.2 Machine learning algorithms for predictive modeling Machine learning (ML) algorithms, such as support vector machines, random forests, decision trees, and artificial neural networks, have been widely used for predictive modeling in drug discovery. These algorithms help in constructing models that predict the biological activity of new ligands, optimize hits, and predict the pharmacokinetic and toxicological profiles of compounds (Lima et al., 2016). ML techniques have also been employed in ligand-based and structure-based drug design studies, including similarity searches, classification models, and virtual screening. The integration of ML with traditional computational methods has improved the prediction of binding sites and docking solutions, thereby enhancing the efficiency of drug discovery. 5.1.3 Case studies of successful AI integration Several successful case studies highlight the integration of AI in drug discovery. For instance, AI has been used to predict drug-target interactions and molecular properties, leading to the identification of potential drug candidates with high accuracy (Han et al., 2023). In another example, deep generative models have been employed to explore chemical space and expedite the drug discovery process by generating novel compounds with desired properties (Born and Manica, 2021). These models leverage multimodal data sources to map biochemical properties to target structures, demonstrating the potential of AI to revolutionize drug design. Additionally, AI-based methods have been successfully applied in virtual screening and de novo drug design, showcasing their ability to handle large datasets and complex molecular interactions (Batool et al., 2019). 5.2 High-throughput virtual screening High-throughput virtual screening (HTVS) has become a cornerstone of modern drug discovery, allowing researchers to rapidly evaluate large libraries of compounds for potential biological activity. HTVS methods, such as molecular docking, pharmacophore modeling, and quantitative structure-activity relationship (QSAR) models, have significantly reduced the time and cost associated with drug discovery (Sliwoski et al., 2014). These methods enable the identification of promising drug candidates by simulating their interactions with target macromolecules and predicting their binding affinities. The integration of AI and ML techniques with HTVS has further enhanced its efficiency, enabling the rapid screening of vast chemical libraries and the identification of potential leads with high accuracy. 5.3 Multi-scale modeling approaches Multi-scale modeling approaches in drug discovery involve the integration of various computational methods to study biological systems at different scales and dimensions. These approaches combine molecular dynamics simulations, quantum mechanics, and coarse-grained modeling to provide a comprehensive understanding of drug binding sites and mechanisms of action (Lin et al., 2020). By bridging different scales, multi-scale modeling allows for the detailed exploration of molecular interactions and the prediction of drug efficacy and safety. The use of AI and ML in multi-scale modeling has further enhanced its predictive power, enabling the accurate simulation of complex biological processes and the identification of potential drug candidates (Keith et al., 2021). 6 Case Studies in Drug Discovery 6.1 Successful drug candidates developed using computational chemistry Computational chemistry has significantly contributed to the development of several successful drug candidates. For instance, the use of computer-aided drug discovery (CADD) methods has been instrumental in the development of anticancer drugs. These methods have provided valuable insights into cancer therapy, making the drug design process faster, cheaper, and more effective (Cui et al., 2020). Additionally, computational tools have been employed to identify novel antimalarial drugs, addressing the urgent need for potent treatments against drug-resistant strains of malaria (Duay et al., 2023). Moreover, the integration of molecular docking and virtual screening techniques has led to the identification of promising drug candidates for various diseases. For example, the construction of a natural product database and subsequent molecular docking experiments have identified potential ligands targeting the androgen receptor for prostate cancer treatment (Huang et al., 2021). Similarly, computational methods have been used to screen active
RkJQdWJsaXNoZXIy MjQ4ODYzNA==