The vast majority of medicines available today are small molecule drugs, composed of just a few hundred atoms or fewer. These drugs range from common over-the-counter remedies like aspirin to advanced targeted therapies used to treat diseases – such as cancer, diabetes, autoimmune disorders, and respiratory infections – and account for 90% of global drug sales. [1]

However, the process of discovering and developing new small molecule drugs is both costly and time-consuming, often taking over a decade and costing upward of $1-2 billion, with a staggering 90% of drug candidates failing in clinical trials. [2]

Fortunately, artificial intelligence (AI) holds immense potential in expediting drug discovery by enabling researchers to swiftly identify the most suitable molecules.

Discovering novel small molecule drugs is a long, iterative process that involves scrutinizing thousands of chemical compounds to identify the most suitable one.

The initial step in the process involves biologists identifying a potential drug target, such as an enzyme that has a critical role in a disease. Medicinal chemists then embark on a quest to discover a small molecule that can modify the target's activity. Apart from inducing the desired biological response, the compound must possess several other vital ‘drug-like’ characteristics, such as non-toxicity, solubility, and stability within the body, to establish its efficacy as a successful drug.

The conventional approach to searching for potential drug candidates involves high-throughput screening, which utilizes automated equipment to rapidly screen vast collections of small molecules to identify potential hits with the desired activity against the target. The next step is to convert these hits into lead compounds that can potentially become effective drugs. This process, commonly referred to as hit-to-lead, involves a series of chemical modifications to the hit molecules to enhance their potency, selectivity, and other drug-like properties.

The hit-to-lead process often involves various iterations of chemical synthesis, biological testing, and computer modeling, requiring significant time, resources, and expertise. Moreover, a considerable proportion of these initial compounds is discarded during subsequent stages of development due to their ineffectiveness or unsuitable drug-like properties or the challenges associated with their chemical synthesis.

“One of the primary challenges in drug discovery is the low probability of identifying the ideal drug-like molecule during the initial stages,” says Ashwini Ghogare, our Head of AI and Automation. “There is a significant risk of spending several months attempting to synthesize compounds that may not be feasible or may lack favorable biological characteristics.”

Traditional computational chemistry methods have played a major role in drug discovery for many decades. For example, computer-aided drug design (CADD) uses structural information about candidate molecules and their targets to predict how they may interact.

However, drug hunters are now turning to alternative strategies to improve the odds of success. AI and machine learning models can extract hidden insights from huge datasets – streamlining the discovery process and saving time and resources by preventing work on compounds that are destined to fail.

“Computational chemists can now deploy ultra-fast methods to search through the universe of virtual chemical spaces containing billions of molecules to find innovative structures with better drug-like profiles,” says Ghogare.

There are an estimated 166.4 billion molecules in the chemical universe database GDB-17. [3] But only a small fraction of chemical space has so far been explored in search of new small molecule drugs – as even these vast virtual libraries barely scratch the surface of all possible structures.  

Our new AIDDISON™ drug discovery software taps into the power of AI, machine learning, and CADD methods – providing a one-stop shop for AI-generative methods, virtual screening of large chemical spaces, and tools for hit-to-lead discovery and optimization.

This AI-powered drug discovery software allows researchers to explore unbounded chemical space and generate ideas for entirely new compounds. AIDDISON™ can help to rapidly identify the best drug-like candidate molecules based on their predicted activity. The tool also includes a synthetic accessibility score from our SYNTHIA™ retrosynthesis software, which predicts whether it will be possible to make the compound through chemical synthesis.

“Integrating these technologies will accelerate drug discovery and find molecules that are much more likely to succeed,” says Ghogare.

But leveraging the power of AI and machine learning in drug discovery relies on the availability of enough clean and relevant data for training these models so that they can provide scientifically valid predictions.

AIDDISON™ is the first commercially available product that includes machine learning models that have been trained on experimentally generated datasets from pharmaceutical R&D.

While most major pharmaceutical companies are developing their proprietary AI-powered drug discovery systems using internal experimental data, we are making these tools accessible to scientists in the broader research community worldwide, including small and medium-sized pharmaceutical enterprises.

“We’ve trained these machine learning models using our proprietary data and built it into the software – to increase the probability of successfully finding the right candidate molecule,” says Ghogare. “Our customers will also be able to train these models using their data and incorporate those learnings as well, enabling them to do even better drug discovery.”

“AIDDISON™ is like a Swiss Army knife,” says Ghogare. “Integrating all of these different technologies into a single platform is a winning formula for faster and cheaper drug discovery.”

By combining traditional computational chemistry with transformative AI and machine learning techniques in one easy-to-use tool, we hope AIDDISON™ will help drug hunters to find the right molecule sooner.

“Bringing together our drug discovery expertise from healthcare and our R&D and customer-facing expertise from life science puts us in a unique position to innovate and build transformational products to accelerate drug discovery,” explains Ghogare. “And ultimately, to deliver benefits for patients.”

^References:

^{[1] https://www.immpressmagazine.com/small-vs-big-understanding-the-differences-between-small-molecule-drugs-and-biologic-drugs/
[2] https://www.sciencedirect.com/science/article/pii/S2211383522000521
[3] https://pubs.acs.org/doi/pdf/10.1021/ci300415d?src=getftr}