In the rapidly evolving landscape of medical science and research, the emergence of artificial intelligence (AI) technologies has proven to be a transformative force in the field of drug discovery. Traditional screening processes and the subsequent analysis of collected data are both time and capital intensive. However, with the rise of AI, there have been significant opportunities to optimise existing processes, or present viable alternatives entirely. This blog post outlines some examples of early-stage drug discovery processes that could either incorporate AI, or are already being explored with AI in mind.
AI In High-Throughput Screening
High-throughput screening (HTS) is a long-established drug discovery process used to assess the biological or biochemical activity of a large number of compounds by testing them against drug targets in parallel using automated systems. The resultant enzyme inhibition, receptor binding, or downstream cellular changes are detected using sensitive methods such as fluorescence and luminescence.
While this process allows for the efficient screening of a plethora of compounds, the subsequent data generated requires thorough analysis of troves of information – a heavy time investment. To assist with this, research is underway to examine whether AI integration into HTS workflows utilising machine learning (ML) derived algorithms can efficiently automate screening analysis to identify hit and lead compounds.
AI in Computational Drug Discovery
More obvious is the potential role that AI can play in improving the area of computational drug discovery. Virtual screening via ligand-based and structure-based methods can be enriched and optimised with the use of AI. For instance, molecular docking has traditionally been a computationally intensive process requiring search algorithms to explore, predict and score the conformations of a ligand to a binding site. However, it is noted that such algorithms sacrifice predictive accuracy for computational speed. With the rise of AI, machine learning-developed algorithms have been shown to vastly improve efficiency while maintaining accuracy.
Another (and arguably the most exciting) example of AI integration in drug discovery is in the nascent industry of generative AI-driven drug design – responsible for the imagination and generation of novel drug candidates. This approach aims to create new chemical entities based only on information relating to their biological target or their known active binders. While computer-driven de novo drug design has been in existence since the 1990s, the rise of generative AI with access to vast databases of protein structures, toxicity profiles, and bioactivity of compounds has allowed novel drug candidates to be designed and developed with greater specificity at a rate never seen before.
Industry Interest in AI
The industry has taken note of the potential of AI in drug discovery, with numerous new companies growing successfully in this area. There is clearly expressed interest from traditional and well-established large pharmaceutical companies in developing capabilities in-house (including AstraZeneca’s ambition to build a data backbone within their business), establishing partnerships with innovative startups (including Recursion’s collaboration with Roche and Genentech), and connecting with leading tech companies (such as Genentech and Amgen’s blockbuster agreement with Nvidia).
In line with both the growing appetite to expand business functions in drug discovery based on AI, and the actual capabilities of AI in this area increasing on a daily basis, there have now been several AI-designed drugs that have made it to clinical trials. These include:
- DSP-1181, developed by Exscientia and Sumitomo for Obsessive-compulsive Disorder
- INS018_055, developed by InSilico Medicine for Idiopathic Pulmonary Fibrosis
- EXS-21546, developed by Exscientia for solid tumours
Caveats and a Conclusion
While AI shows promise in accelerating early-stage drug discovery, caution is necessary to temper expectations. Critics highlight the need for independent verification of AI-generated drugs and caution against overreliance on AI due to its limitations in generating realistic data. Furthermore, clinical validation of the effectiveness of any AI-designed molecule is the same as any other identified candidate and would be the most significant driver of expenditure in the drug development pathway.
Nevertheless, the potential for AI to shift paradigms is undeniable. With advancements in algorithms and significant industrial collaborations, AI can streamline and enhance the early-stage drug discovery process. By identifying promising candidates efficiently, it promises faster and more cost-effective treatment development, offering hope for breakthroughs in healthcare.
By harnessing advanced AI-driven technologies, it may be possible to streamline the identification of promising compounds for use in future therapeutics, revolutionising traditional processes seen in early-stage drug discovery research.
Photo credit: Image design by Freepik.
