Using biological target structure in computer-aided drug design

The world of pharmaceutical research is being revolutionised by recent advances in computational techniques and computer power [Khan2024]. Encompassing sophisticated computational methods to design and optimise new drug candidates, Computer-Aided Drug Design (CADD) lies at the core of this transformation and is becoming an integral part of how new drugs are discovered [Sabe2021]. Here at Oxcitas, we are using CADD to make drug development faster, more targeted, and more efficient.

By combining molecular modelling, bioinformatics, and cheminformatics, CADD approaches can predict how drugs will interact with their biological targets, allowing the visualisation and analysis of crucial molecular interactions that are key for a drug’s effectiveness and safety.

Structure-Based Drug Design

One important part of CADD is Structure-Based Drug Design (SBDD) [Batool2019]. SBDD focuses on understanding the detailed three-dimensional structure of target proteins to identify and refine small molecules that can bind to these proteins and change their activity. This approach involves many small steps, often requiring various iterations, but gives valuable insights into protein-ligand complexes and helps in designing more potent and selective drug candidates.

Several successful drugs have come out of SBDD approaches, highlighting their potential in this field. One of the first successes was the development of HIV-1 Protease Inhibitors [Wlodawer1998] which proved critical in the fight against HIV/AIDS during the 1990s and greatly improved patient outcomes. Another notable success is STX-0119 a small molecule inhibitor which targets signal transducer and activator of transcription 3 (STAT3), a protein involved in various cancers, including lymphoma. The development of STX-0119 was advanced using structure-based virtual screening (SBVS) [Matsuno2010].

SBDD relies on having high-quality three-dimensional structures of biological targets, typically proteins. Protein structures are usually obtained through techniques like X-ray crystallography, Nuclear Magnetic Resonance (NMR), and Cryo-Electron Microscopy (Cryo-EM). Alternatively, structures can be generated computationally by homology modelling or AI-based methods. Even though the reliability of computationally generated protein structures is still below what can be observed experimentally, virtually generated  structures are very valuable for initial explorations and for cases where experimental data is simply not available yet. 

Often choosing the right structure can be challenging due to protein conformational variability and structure quality, and it is important to consider a number of quality metrics as well as the particular goal of the study before selecting the most-suitable representation for the analysis. Drug-target complexes (i.e., protein structures bound to known small-molecule compounds) are especially valuable as they provide insights into the binding site location, shape, and composition. This information is crucial for designing molecules that interact precisely with the target’s active sites, improving the drug’s effectiveness and safety. Sometimes, if the target is known to interact differently with a diverse set of molecules, ensembles of representative structures can be used to cover multiple relevant scenarios [Wang2017].

Once the target reference structure - or structures - have been selected, a large set of drug candidates is considered and virtually screened against said target in order to identify potential hits. The specific library of compounds used for screening is dependent on the ultimate goal of the analysis but generally must pass specific physico-chemical filters as well as meeting specific criteria for a first elimination of undesirable properties. Proper preparation of ligand structures for successful virtual screening is also crucial.

Dedicated computational tools are used to predict how compounds fit into the target protein’s binding site, focusing on their bioactive conformation and interactions with the protein. Molecular docking is a popular technique for exploring chemical spaces efficiently and the construction of increasingly more effective AI/ML-based approaches is possible due to the growing availability of large and more heterogeneous datasets on drug-target interactions. Recently, AI and machine learning (ML) techniques have advanced the prediction of compound binding affinity and modes, often achieving greater accuracy than traditional methods. 

The final element of SBDD revolves around optimisation of the identified hits. This process involves improving their drug-like properties, including potency, selectivity, and ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) profiles. This process also addresses issues like Cyp liabilities (important for any combination therapy), drug resistance mechanisms, and interactions with the human ether-a-go-go (hERG) potassium channel to avoid potential cardiotoxicity risks. This iterative refinement ensures that the drug achieves the desired pharmacological effects.

Future Directions

SBDD is demonstrating the power of combining structural biology with computational chemistry, to compress the costs and times associated with traditional wet-lab approaches to drug discovery, thereby offering a promising cost-effective alternative path. The future of SBDD looks bright. With advances in computational power, AI, and machine learning set to enhance prediction accuracy and drug design efficiency, CADD approaches such as SBDD will likely open up new possibilities for discovering novel and exciting therapeutic options across a range of diseases [Sadybekov2023].

References

[Khan2024] Khan MK et al. (2024) The recent advances in the approach of artificial intelligence (AI) towards drug discovery. Frontiers in Chemistry 12:1408740.

[Sabe 2021] Sabe VT et al. (2021) Current trends in computer aided drug design and a highlight of drugs discovered via computational techniques: A review, European Journal of Medicinal Chemistry, 224, 113705.

[Batool2019] Batool M, Ahmad B, and Choi S (2019). A Structure-Based Drug Discovery Paradigm. International Journal of Molecular Sciences, 20(11):2783.

[Wlodawer1998] Wlodawer A and Vondrasek J (1998) Inhibitors of HIV-1 protease: A major success of structure-assisted drug design. Annual Review of Biophysics and Biomolecular Structure, 27, 249–284

[Matsuno2010] Matsuno K et al (2010). Identification of a New Series of STAT3 Inhibitors by Virtual Screening. ACL Medicinal Chemistry Letters, 1(8): 371–375.

[Wang2017] Wang N, Wang L, and Xie X-Q (2017) ProSelection: A Novel Algorithm to Select Proper Protein Structure Subsets for in Silico Target Identification and Drug Discovery Research. Journal of Chemical Information and Modeling, 57 (11), 2686–2698

[Sadybekov2023] Sadybekov AV and Katritch V (2023). Computational approaches streamlining drug discovery. Nature 616, 673–685.