Drug screening is the initial link and key to drug development. In the past, traditional drug screening methods such as animal models were often costly and time-consuming. With the development of technology and the improvement of the overall level of computational chemistry, the application of computer chemistry-related technologies to perform virtual screening has become a reality. Virtual screening is a new drug screening method developed with the help of computer technology based on the lock-and-key theory between targets and drug molecules. It can use professional computer software to simulate the binding of ligands and corresponding receptors and select seed compounds from the compound database by calculating the binding ability between the drug and the corresponding target. The use of virtual screening methods to discover leads greatly enhances the pertinence of drug research, increases the hit rate of biological activity tests, and thus reduces the consumption of experimental costs.
In principle, virtual screening can be divided into two categories, namely, receptor-based virtual screening and ligand-based virtual screening.
Receptor-based virtual screening starts from the three-dimensional structure of the target protein, studies the characteristic properties of the binding site of the target protein and the interaction mode between it and small molecule compounds, and evaluates the protein and protein according to the affinity scoring function related to the binding energy. The binding ability of small molecular compounds is evaluated, and finally, compounds with reasonable binding modes and high prediction scores are selected from a large number of compound molecules for subsequent biological activity tests.
Ligand-based virtual screening generally uses small molecule compounds with known activity to search for chemical molecular structures that can match it in the compound database according to the compound’s shape similarity or pharmacophore model. Finally, the selected compounds were screened experimentally.
Compound Libraries Commonly Used in Virtual Screening
Classification of screening compound libraries: diversity compound library, marketed drug molecule library, known activity library, target compound library, natural product library, fragment library, etc.
Preprocessing of Compound Libraries
In the early stage of drug discovery, accurate and rapid exclusion of non-drug compounds will help to enrich active compounds and reduce screening costs. In the compound sample preparation stage of drug screening, the removal of non-drug-like compounds is often the first method adopted. It is based on the characteristics of the drug-likeness of the compound and excludes the compounds in the compound database that violate the drug-likeness attributes of the compound. This method is simple and easy and can be completed in the compound library management system. Non-pharmaceutical compounds mainly include the following types:
- Compounds with non-drug elements such as transition metal elements;
- Compounds with a molecular weight less than 100 or greater than 1000;
- Compounds with a total of less than 3 carbon atoms;
- Compounds without nitrogen, oxygen, or sulfur atoms;
- Compounds that violate two or more of the “Lipinskis rule of five”;
- For non-central nervous system drug screening, compounds with blood-brain barrier coefficient logBB greater than 03 should be excluded. Where logBB is the logarithm of the steady-state concentration ratio of the drug molecule in the brain and blood, ie log(C/C);
- For drug screening of the central nervous system, compounds with blood-brain barrier coefficient logBB less than 0 should be excluded.
Commonly Used Virtual Screening Software
Molecular docking, as the most important method in structure-based virtual screening, can predict the binding mode of target and ligand. In the past 20 years, a large number of molecular docking software and programs have been developed, including Autodock, Autodock Vina, LeDock, rDock, UCSF DOCK6, LigandFit, Glide, GOLD, MOE Dock, Surflex-dock, etc.
For molecular docking software, the two most critical parts are the sampling algorithm and the scoring function, which respectively determine the sampling and scoring capabilities of the software. Currently, popular sampling algorithms can be roughly divided into three categories: shape matching, systematic search (exhaustive search, segmentation, and conformation ensemble), and random search algorithms (such as the Monte Carlo algorithm, genetic algorithm, tabu search method, and group optimization method). The popular scoring functions can be mainly divided into three categories: force field, experience, and knowledge-based scoring functions.
Table 1: Summary of information on commonly used virtual screening software
References
1. E F. Influence of Configuration on the Action of Enzymes. Journal of the American Chemical Society, 1894, 27: 2985-2993.
2. Hawkins, P.C.D., Skillman, Nicholls, A., J.Med. Chem., 2007, 50, 74.