AI-assisted Virtual Screening

Techniques for virtual screening (VS) have grown in importance as a tool for lead discovery. In order to identify subsets of a database or library that are more manageable for the subsequent experimental high throughput screen (HTS), VS method can screen out molecules that have steric clashes with active site residues or that exhibit polarity property that clash with the active site electrostatic characteristics. This results in significant time and costs savings. Various types of drug-like or lead-like molecular libraries are calculated and screened in VS based on the well-known 3D structures of target proteins. The filtering of molecule libraries is provided by the docking approach, where compounds are evaluated based on their binding affinity. VS techniques offer a fast and economical way for the discovery of new actives, as opposed to experimental methods. The two most popular strategies for VS are ligand-based methods and structure-based methods (docking). Creative Biolabs has established the AI-assisted in silico bioinformatics platform to support our customers in molecular docking, molecular dynamics simulation, virtual screening, and quantitative structural activity relationship (QSAR) analysis.

Fig.1 Virtual Screening. (Gentile, et al., 2022)

Structure-based Virtual Screening

Early-stage drug development focuses on the identification of lead compounds that have pharmacological activity against a biological target and the progressive optimization of the pharmacological characteristics and potency of these compounds. Structure-based drug discovery (SBDD) is a vital method for assisting fast and cost-effective lead finding and optimization. Rational structure-based drug design has proven to be more effective than traditional drug discovery approaches because it aims to understand the molecular basis of disease and in the process leverage knowledge of the three-dimensional structure of biological targets. Today, it allows us to explore the underlying molecular interactions involved in ligand-protein binding and interpret experimental results in atomic-level detail by using computational methods and the 3D structural information of the protein target. Processing the 3D target structural information of interest, identifying the binding site, and preparing the compound database are the first steps in the general workflow of an SBVS method. Following library and receptor preparation, each compound in the library is virtually docked into the target binding site with a docking program. By investigating the conformational space of the ligands inside the binding site of the protein, docking aims to predict the structure of the ligand-protein complex. The free energy of binding between the protein and the ligand in each docking pose is then roughly estimated using a scoring algorithm. Following docking and scoring, ranked compounds are post-processed by the validity of generated pose, examining calculated binding scores, metabolic liabilities, undesirable chemical moieties, lead-likeness, desired physicochemical properties, and chemical diversity.

Fig.2 Workflow for the discovery of mutant-specific PI3Kα inhibitors based on a SBVS protocol. (Lionta, et al., 2014)

Ligand-based Virtual Screening

The ligand-based virtual screening (LBVS) is based on the assumption that molecules with similar structures tend to have similar properties and functions. Using the chemical and physiochemical similarities of active ligands, LBVS can predict the other active ligand from a pool of compounds with high bioactivity and attempt to prioritize candidate molecules. This technology is a useful alternative for hit identification when the 3D structure of the target is unknown or it is difficult to carry out virtual drug screening using structure-based approaches. Typically, a simple LBVS process consists of only a few steps. Each of the molecules used as input is first given a molecular representation. Second, an evaluation of similarities between potential compounds and known active molecules will be done. The candidates are then ranked in accordance with their individual scores in the next stage. Finally, from a library of many inactive compounds, a limited number of active compounds will be found. Working in the area of virtual screening for years, Creative Biolabs has developed state-of-the-art software for ligand-based virtual screening. The LBVS approaches include quantitative structure-activity relationship modeling, similarity searching, and pharmacophore searching.

Fig.3 Ligand-based pharmacophoric features of G-1. (Khan, et al., 2019)

References

1. Gentile, F.; et al. Artificial intelligence–enabled virtual screening of ultra-large chemical libraries with deep docking. Nature Protocols. 2022, 17(3): 672-697.
2. Lionta, E.; et al. Structure-based virtual screening for drug discovery: principles, applications and recent advances. Current topics in medicinal chemistry. 2014, 14(16): 1923-1938.
3. Khan, S.U.; et al. Sequential ligand-and structure-based virtual screening approach for the identification of potential G protein-coupled estrogen receptor-1 (GPER-1) modulators. RSC advances. 2019, 9(5): 2525-2538.