Computational Methods in Drug Discovery

By Sarah MooreReviewed by Emily Henderson, B.Sc.

The drug discovery and development process is incredibly costly and time-consuming. Recent figures estimate that by the time a new drug gets market approval, $2.6 billion may have been spent on the process. Much of this cost is attributed to spending time researching candidate drugs that fail to pass phase I trials, with up to 90% of candidate drug molecules identified in the discovery stage failing to get to market.

Drug Discovery. Image Credit: Gorodenkoff/Shutterstock.com

Computational methods have emerged as an approach that is revolutionizing the face of drug discovery. Techniques of computer-aided drug discovery (CADD) can significantly reduce the time and money spent at the early stage of drug discovery by intelligently highlighting molecules that have a strong chance of acting in the desired way on the drug target without inducing reverse adverse events.

In modern drug discovery, CADD tools have become an invaluable tool. As computational power and capabilities along with our knowledge of biological systems have grown, the applications of computational methods at various stages of the drug discovery and development pipeline have developed.

Particularly, over the last few decades, several computational drug discovery methods have been widely adopted by the pharmaceutical industry to benefit the drug discovery process, including de novo design, molecular docking, molecular similarity calculation, and sequence-based virtual screening. Here, we discuss these methodologies in detail.

Methodologies of computational drug discovery

De novo design

The de novo design technique has already proven itself successful at aiding the development new of molecules for novel therapeutics in a fast and efficient manner. The method involves leveraging the three-dimensional structure of the target receptor to create new molecules. The process relies on the structural determination of the target complex to generate new molecules that match the target binding pattern.

The most commonly used software solutions for running structure-based de novo designs include LUDI, BUILDER, and CAVEAT. The algorithms designs by these solutions develop novel molecular entities based on potential ligand-receptor interactions that they identify.

Molecular docking

The method of molecular docking involves the investigation of how two or more molecules, such as enzymes, drugs, or proteins, fit together. Molecular docking uses computational methods to predict how an enzyme is likely to interact with a ligand. Two distinct steps are involved in the process of molecular docking.

First, algorithms are used to sample conformations of the ligand within the protein’s active site. Next, the scoring function ranks these confirmations. Best practice molecular docking uses sampling algorithms that reproduce the experimental binding mode, in addition, the experimental binding mode should be ranked highest out of all generated conformations by the scoring function.

Molecular similarity calculation

The SHApe-FeaTure Similarity method, or SHAFTS, was developed based on the pharmacophore matching approach. The SHAFTS method was created to establish rapid 3D molecular similarity calculation.

The SHAFTS method utilizes similarity metrics of molecular shape and colored chemistry groups that have been highlighted by pharmacophore features for 3D calculation. It then ranks molecules combined strength in terms of their volumetric similarity approaches and pharmacophore matching. Additionally, the molecular similarity calculation can leverage the triplet hashing method to compute fast molecular alignment poses. Pharmacophore feature fit values are combined with shape-densities overlaps to score and rank alignment.

The SHAFTS method has demonstrated its superior performance in its ability to gain fine-grain data of actives and chemotypes in comparison with ligand-based methods.

Sequence-based virtual screening

Given the fact that for most proteins, scientists have yet to determine their 3D structures and ligands for some proteins are unknown, structure-based methods nor ligand-based methods are appropriate for drug discovery and development of these proteins.

The pharmaceutical industry has needed a reliable and effective method for predicting ligand-protein interactions (LPIs) when structure or ligand information is not known. To meet this need, a sequence-based drug design model for LPI was recently developed. This method uses the support vector machine (SVM) approach and is based on the primary sequence of the target proteins and the structural features of small molecules.

The future of the drug discovery and development process

Although much investment has been made in the area of drug discovery and development in recent decades, there is still a need for faster methods that cut the costs and time associated with moving candidate drugs through to market approval.

Computational methods are currently offering a solution, leveraging the evolution of computational technology to speed up and streamline the drug discovery process by highlighting drugs with good chances of meeting functional requirements and eliminating unsuitable candidate molecules early on without wasting resources putting them through clinical trials.

Computational methods of drug discovery will likely continue to be a point of focus for the pharmaceutical industry in the coming years, with more investment expected in the area to push capabilities further.

Further Reading

• All Drug Discovery and Manufacturing Content
• Drug Discovery, Manufacturing, and Development
• What are Peptide Decoys?
• Using Artificial Intelligence to Discover New Antibiotics
• Applications of 3D printing in Drug Delivery