Structure-Based Drug Discovery & Design Services : Strategies and Implementation

Basic Strategy 1: Known Ligand and Protein Structures

This strategy, shown in the blue quadrant (image above), offers the most straightforward approach, leveraging the known structures of both the ligand and the protein target to enable efficient structure-based drug design. Computational methods, such as QSAR (Quantitative Structure-Activity Relationships) modeling, are commonly used to predict novel active molecular structures at the initial stages of a drug discovery project. These models analyze relationships between chemical structures and biological activity by using basic molecular features such as charge distribution, hydrophobicity, and stereochemistry of known ligands. In addition, the structures of the ligand-protein complexes enable the construction of a pharmacophore model. Pharmacophore modeling uses essential 3D features to map the interactions between a known compound and a target protein, including the arrangement of ligand-binding atomic groups (side chain and main chain atoms) and the distribution of charged and hydrophobic groups within the binding site. The model can subsequently guide the design of new active molecules. Both QSAR and pharmacophore modeling are used in virtual screening of fragment or drug-like compound libraries to improve the efficiency and accuracy of the method. An advantage of virtual screening is its ability to handle much larger compound libraries and a higher screening speed. In recent years, AI-based screening methods have gained popularity. In a blog article, you can find an example of the use of AI in fighting antibiotic resistance.

Biophysical screening techniques such as NMR spectroscopy, X-ray crystallography, and our proprietary weak-affinity chromatography (WAC™) play a central role in compound library screening and identification of new active compounds. Recent developments in synchrotron radiation technologies have made it possible to use X-ray crystallography in fragment screening. The advantage of the crystallographic fragment screening method is that it provides direct insights into the binding mode of fragment hits. NMR spectroscopy is highly efficient for detecting low-affinity ligands (Kd > 10−6 M) or for studying the protein in a non-ideal conformation for ligand binding using crystal soaking.

The Design-Make-Test-Analyze Cycle

The activities of the new compounds identified in screening campaigns need to be verified using biochemical and cell-based assays. The best compounds are then taken to the next optimization cycle. The process of optimization, often designated as the design-make-test-analyze cycle, can involve many cycles of synthesis-activity assessment, crystallization, and structure determination, followed by the use of structural data to guide the synthesis of new compounds. During these cycles, multiple protein-ligand complex structures are determined, e.g., by X-ray crystallography, NMR spectroscopy, or cryo-EM. The large number of ligand-protein complex structures determined during these cycles, as well as during crystallographic fragment screening, requires access to synchrotron beamlines with streamlined data collection and processing facilities. The beamlines at the MAX IV synchrotron in Lund, BioMax, and MicroMax, located just a few kilometers from SARomics Biostructures’ laboratories, are well-suited for these aims.

Basic Strategy 2: de novo Design: Known Protein Structure, Unknown Ligand

In the green quadrant of the image above, we focus on Basic Strategy 2: cases where the protein structure is known, but no known ligand structure is available. Here, we need to conduct fragment or compound library screening at the start of a structure-based drug design. Since most proteins currently have AlphaFold-generated 3D structures, virtual screening can also be applied to these models after constructing QSAR and pharmacophore models. In addition, as in the Strategy 1 case, we can use biophysical screening methods like NMR spectroscopy or weak-affinity chromatography (WAC). Once we identify suitable binders, we can enhance our QSAR model and also clone, express, purify, and crystallize the protein in complex with the best ligands, moving us toward Basic Strategy 1. This approach validates the binding mode, offers direct insights into stabilizing interactions within the protein-ligand complex, and aids structure-based discovery and design. We should also note that having a ligand at the protein’s binding site often makes crystallization of the complex easier, which might otherwise be difficult. If we cannot obtain a crystal, we can continue with our QSAR and pharmacophore models, use, e.g., NMR spectroscopy to verify binding modes, and perform biochemical and cell-based assays to confirm activity. In other words, we continue to use a pure ligand-based drug design strategy (Basic Strategy 3).

Basic Strategy 3: Ligand-Based Drug Design (LBDD)

The yellow quadrant in the image above relates to Basic Strategy 3, ligand-based drug design (LBDD). In this case, the ligand’s structure is known (e.g., a substrate or inhibitor), but the protein target’s experimental structure is unavailable. The principle of this method suggests that knowing the structure of one or more ligands that bind to a protein allows us to create additional ligands with similar structures and equal or greater binding affinity. In this scenario, we can, for example, use the protein’s AlphaFold model to obtain details of the protein’s ligand-binding site, and use the known ligand (or ligands) structure to build a pharmacophore model based on the ligand’s shape and the presence of charged, polar, and hydrophobic groups. When several known ligands are available, their common features can be overlaid to improve the accuracy of the pharmacophore model.
We can also use a computational method called ligand hopping. In this case, computational chemistry, machine learning, and AI methods, together with medicinal chemistry expertise, are employed to identify new chemical structures with similar activity. The ligand-hopping method in drug design is often used to obtain a new structure (chemotype), e.g., to overcome patent protection, improve the characteristics of a known ligand, or even to rescue failed leads with some degree of activity.

The AlphaFold model in combination with the pharmacophore model can also be used, e.g., to perform virtual screening of a chemical library to identify similar compounds and to conduct quantitative structure-activity relationship (QSAR) analysis. Subsequently, the hits are ranked, and the best binders are selected for biochemical and cell-based assay analysis. This process is repeated until good-quality lead compounds are obtained. It should be noted that computational methods, even when combined with an AlphaFold model, may yield inaccurate results.

Ligand-Based Drug Design Challenges & the Benefits of SBDD

The lack of experimental 3D structures in LBDD presents several challenges. One major limitation, as noted above, is that this method relies heavily on high-quality ligand data and does not provide direct insights into ligand binding to their targets. Consequently, this can significantly slow down the optimization of initial hits. In contrast, traditional SBDD offers direct mechanistic insights into ligand-binding interactions and facilitates the rational design of new ligands. This explains the benefits of structure-based drug design. LBDD can be useful in the early stages of a project, while the target’s experimental structure and its ligand complexes are being determined using traditional techniques such as X-ray crystallography, NMR spectroscopy, or cryo-electron microscopy (cryo-EM). An experimental 3D structure in complex with a ligand will enable the implementation of Basic Strategy 1, ultimately accelerating the drug discovery process.

Basic Strategy 4: de novo Design: Unknown Ligand, no Experimental Protein Structure

Even in the absence of a known ligand and an experimental protein structure, we can still use virtual screening with an AlphaFold 3D model of the protein. However, before that, we need to determine the protein’s drugability and identify a potential ligand-binding site. Assuming that we have access to a purified protein, we can also run a high-throughput screen (HTS) in combination with a biochemical assay or use NMR spectroscopy to screen a fragment library.

If the target protein’s experimental structure is unavailable, SARomics Biostructures provides custom gene-to-protein-structure services. In such cases, the protein is cloned, expressed, purified, and crystallized, and its X-ray crystallographic structure is determined. In addition, our library of FastLane structures is designed to accelerate our clients’ projects by offering over 600 proteins for co-crystallization with the client’s ligands, with a turnaround time of a few weeks.