Virtual Screening Techniques in Drug Discovery

By Sarah MooreReviewed by Emily Henderson, B.Sc.

Drug discovery in the pharmaceutical industry is hindered by the huge costs involved in searching for new therapeutics. It is estimated that a new drug costs an average of $2.6 billion to develop, with much of that money wasted on researching candidate drugs that fail to get approved past phase I trials.

Drug Discovery. Image Credit: PopTika/Shutterstock.com

Approximately nine out of ten candidate drug molecules are unsuccessful. This high rate of failure contributes to billions of dollars and significant working hours wasted each year, making drug discovery inefficient and expensive.

Now, the pharmaceutical industry’s largest players are adopting new technology that promises to make the drug discovery process quicker, cheaper, and more efficient.

What is virtual screening?

Virtual screening (VS) is a computational technique that is being implemented by pharmaceutical companies to improve their research and design phase. VS automatically searches through libraries of molecules and selects those that have structures considered to be most likely to effectively bind to a therapeutic target, such as a protein receptor or an enzyme.

VS is one of the numerous emerging strategies pharmaceutical companies are exploring to streamline their drug discovery processes. Over recent years, VS has demonstrated its efficacy as a strategy for effectively identifying bioactive molecules and presents the pharmaceutical industry with the potential to drastically speed up the drug discovery phase.

There are two main strategies implemented by VS, ligand-based screening techniques and structure-based screening techniques.

Ligand-based screening

Pharmacophore models exploit the collective information contained in the structurally diverse set of ligands that bind to a given receptor, allowing a model of the receptor to be built from such information.

The pharmacophore model generated from this collective information is used to compare against candidate ligands, to identify those that are compatible and therefore likely to bind. This approach is the most commonly used ligand-based VS approach. It identifies molecules that have a high likelihood of fitting with the target’s binding site, reducing the amount of time and money spend investigating candidate molecules that structurally will not fit the target site.

Additionally, ligand-based screening approaches can take the form of 2D chemical similarity analyses which are used to scan libraries of molecules and compare them against one or multiple active ligand structures.

The method of ligand-based VS takes just a fraction of a second to run a comparison between a candidate molecule and a target structure, meaning that thousands of potential therapeutic molecules can be investigated in a short timeframe. Researchers require just a single computer processing unit (CPU) to perform such screening techniques, making the process simple and accessible to most labs.

Structure-based screening

The second main approach to VS in drug discovery is structure-based screening which involves the docking of candidate ligands to a target protein. These dockings are then scored automatically to demonstrate the likeliness that the ligand will bind to the target protein with high affinity.

Unlike ligand-based screening, the structure-based approach requires more than a single CPU. The structure-based approach relies on a parallel computing infrastructure to manage large datasets coming in from significant libraries and run multiple comparisons in parallel. High-speed indexing engines are preferred over commercial database engines for carrying out this task.

The future of virtual screening in drug discovery

In recent years, VS has emerged as a groundbreaking technique that is helping to significantly improve and speed up the process of drug discovery. Research has shown VS to be effective at simultaneously scanning the potential affinity of millions of compounds to selected targets. As a result, VS has increased the number of candidate drugs and has helped pharmaceutical companies get these potential candidates into clinical trials far quicker than they would have done with previous methods.

Machine learning (ML), a subset of artificial intelligence (AI), has been fundamental to the success of VS. When applying ML to VS, scientists use known active compounds and inactive compounds to create a training set to tech a model. Once the model has been trained and validated, it can be used to process new compounds and identify those that have the required drug target binding activity.

Already, the pharmaceutical industry’s key players have successfully implemented ML programs into their VS techniques. The use of ML has been proven to increase the accuracy of predictions and to significantly reduce the failure rate in comparison with conventional methods. With the help of ML, researchers can quickly screen out candidate molecules that may fit a target but will eventually be rejected for their toxicity. It allows research to cut out candidates that will not make it past pre-clinical trials, searching for new effective pharmaceutical compounds faster, cheaper, and more efficient.

In the near future, we will likely see the use of ML in VS techniques more often in drug development strategies as the industry continues to strive to enhance its research and development processes to get new therapeutics to those that need them in shorter timeframes.

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