Power of Supercomputing and AI: Revolutionizing Allosteric Drug Discovery

January 5, 2024

Commentary contributed by Xavier Barril, Ph.D., Chief Scientific Officer of Gain Therapeutics 

January 5, 2024 | Driven by advancements in technology and data accessibility, computational methods and artificial intelligence (AI) have seen remarkable growth this past decade. This surge has ushered in a new era of problem-solving across a plethora of industries and domains. In particular, the field of drug development is undergoing significant transformation as a result of these advancements, offering solutions to longstanding challenges encountered by the industry. 

Unmasking allosteric-binding sites on the 90% of currently deemed “undruggable” proteins is one area that’s tapping into supercomputing for drug target discovery. By targeting and binding a small molecule to an allosteric site on the protein (a site other than the active, functional site), allosteric modulation can trigger a conformational change in the protein's structure, creating therapeutic possibilities when it is implicated in disease. This conformational change can either enhance (positive allosteric modulation) or inhibit (negative allosteric modulation) the protein's activity to either correct or disrupt its function, and potentially other molecular changes, like stabilizing the protein, that could be beneficial in targeting misfolded proteins.  

The Impact of Computational Methods and AI in Drug Development 

While high-throughput screening (HTS) has been a cornerstone of conventional drug discovery, it has its limitations. The process is costly due to the screening of a large number of compounds and the management of extensive compound libraries. HTS can also yield false positives, for instance when compounds in the library interfere with the assay through unspecific mechanisms. It can also miss potential hits and, because researchers follow up on the most potent hits, they risk missing allosteric modulators that would fare better under physiological conditions.  

HTS also does not inform about the mechanism of action of the hits, necessitating resource-intensive follow-up validation and optimization to develop a drug. Even in the most compact format (1536-well plates), it is estimated that screening one million compounds requires about four tons of plastic that cannot be recycled because it has been contaminated. 

In contrast, emerging technologies that leverage supercomputing, molecular simulations and AI to enhance compound selection, make the drug discovery process more efficient and cost-effective than the HTS process. Couple that with proprietary druggability analyses, it opens the targets and mechanisms of action that were previously deemed "undruggable," i.e. inaccessible for drug development due to their complexity. With this new set of computational tools, researchers can now analyze large datasets and uncover intricate patterns and relationships that were previously elusive, unveiling new drug targets. 

Supercomputing and AI have the potential to dramatically reduce drug development time, which often spans over a decade. With the aid of advanced computational methods, researchers can expedite multiple stages of drug development, from target identification to lead optimization. This acceleration translates to faster access to life-changing medicines for patients. 

Additionally, supercomputing and AI can significantly reduce drug development cost. The cost of developing a new drug averages around $2.8 billion, 40% of which is attributed to the discovery phase, making it a financially intensive process. Computational methods have streamlined drug discovery by optimizing compound selection, reducing the need for extensive experimental testing, and minimizing resource waste. These efficiencies result in cost savings for pharmaceutical companies and, ultimately, for payers and patients. 

AlphaFold: A Turning Point in Protein Drug Discovery 

At the cusp of the AI revolution for protein drug discovery was the launch of AlphaFold, a system developed by DeepMind. The tool predicts the 3D structure of proteins with outstanding accuracy.  

Inherently, structural data is often lacking or not available concurrently with experimental functional data in various biological domains. The AlphaFold Database offers an unprecedented opportunity for structural biologists, who rely on 3D protein structures to explore fundamental biological questions and understand the relationship between a protein's structure and its function.  

It provides 3D protein structure predictions for whole genomes, aiding in the rational design of drugs. A whole host of applications to guide experimental validation and therapeutic development are being developed around AlphaFold, such as the identification of dynamic regions in proteins, or the prediction of how mutations alter protein structures. 

Shaping the Future of Drug Discovery 

AlphaFold is a first tangible demonstration of the potential of AI to disrupt drug discovery. Of course, these developments do not happen in isolation, and the astounding advances seen throughout the information technologies and biotechnology sectors act synergistically, inevitably leading to exponential gains in speed and efficiency of the drug discovery process, dramatically reducing development timelines and streamlining costs. At the same time, they will unlock previously inaccessible drug target and non-standard mechanisms of action, such as allosterism. 

But, while supercomputing and AI play pivotal roles, it's essential to acknowledge that all drugs— but particularly allosteric drugs, with their subtle modulation of biological processes—require close integration with biological experiments to ensure their effectiveness. By combining the power of AI-driven predictions with experimental validation, researchers can usher in a new era of drug discovery, offering hope to patients worldwide. 


With more than two decades of experience in computational chemistry and drug discovery, Xavier Barril, Ph.D., has served as Chief Scientific Officer of Gain Therapeutics since January 2018. Co-authoring more than 90 papers and 14 patents, Barril is the inventor of the innovative platform technology that Gain Therapeutics is exploiting to identify novel allosteric modulators. He can be reached at xbarril@gaintherapeutics.com.