Over the last 10 years, drug developers have increasingly utilized artificial intelligence (AI) in small molecule drug development and manufacturing. From applying machine learning (ML) in drug discovery to establishing AI tools at the core of each step of the development process, pharma organizations are integrating AI into small molecule development with the aim of accelerating timelines.
Despite the benefits that in silico techniques can bring to small molecule discovery, it is important to understand the potential of AI alongside its limitations in this context. In this article, Petra Dieterich, Head of Scientific Leaders, and Francisco Velazquez, Senior Director, Chemistry at Abzena, explore what is possible for the pharma industry when AI is incorporated into small molecule development with a crucial synergistic mix of computational and human expertise.
Intertwining AI into small molecule drug production
Rapid advances in computing power have opened the doors to the possibility of applying rigorous simulation and optimization techniques to small molecule drug discovery, development, and manufacturing. From discovery to commercial manufacture, the pharmaceutical industry has found ways to incorporate AI, with the intent that these tools will improve and accelerate processes.
Over the last 10 years, the predominant focus of the pharma industry when introducing AI to small molecule development has been on drug discovery and the identification of lead candidates. It has been reported that over 150 small-molecule drugs have been discovered by biotech companies using an AI-first approach, and more than 15 of these molecules have already entered clinical trials1.
In silico screening has been applied to vast chemical libraries, using algorithms to identify viable candidates offering the desired attributes while offering increased speed of discovery. As a result, therapeutically promising drug candidates can be identified from a library of billions in mere minutes2.
Advancements in ML have progressed to a point where algorithms can use data and experience to teach themselves, further truncating timelines to the nomination of the drug development candidates. It has been reported that ML can reduce the timeline of a typical drug discovery program from 5 years to just 12 months2.
Incorporation of AI has also been seen in other areas of drug analytics, with software capable of predicting potential toxicity by calculating the binding affinity of ligands3.
As well as in drug discovery, examples of the pharma industry incorporating AI tools are being seen in many different processes involved in small molecule production. One research group has used AI to direct robotic platforms for flow synthesis of organic compounds. Others have employed AI to predict and analyze the impact of potential changes in process parameters such as the time tablets spend under the spray zone in the coating process4,5.
Potential of AI in the small molecule space
Between 2021 and 2030, the use of AI in the global Pharmaceutical Market is predicted to grow at a compound annual growth rate (CAGR) of 29.4%, from $905 million to $9,241 million6. When examining the potential benefits that can arise from introducing AI to small molecule drug production, it is no surprise why this market is predicted to expand so rapidly. The advantages of AI include:
Accelerating speed to market
Applying AI as a tool in drug discovery has the potential to drastically reduce the time taken to identify and rank molecules to progress to synthesis and testing. Each of these candidates should be of higher quality than those identified by traditional methods. Timelines to deliver promising molecules to clinical trials are being significantly reduced, allowing essential medicines to reach patients sooner.
With the potential to identify the most likely small molecule candidates for success as well as predict and analyze manufacturing optimal parameters, AI techniques will minimize risks and ultimately reduce costs. By selecting molecules of high quality with the help of carefully written algorithms, the chances of success are likely to be higher, reducing the risk of losing both money and time by progressing candidates that may prove suboptimal during later phase development studies
Expanding the patient population
By reducing the costs and time involved in drug discovery, design, and development, AI also has the potential to expand the treatable patient population. Disease areas currently suffering from little investment due to poor return, including orphan diseases or territories in developing countries, could witness an influx of accessible medicines as a result of development cost reductions brought about by AI.
Understanding the limitations of AI
Despite the myriad of potential applications and benefits, a theme amongst some drug developers has been a reluctance toward the uptake of AI. Possibly driven by the fear of job security, it is important to remember that AI should be used as a tool to expand and elevate what drug developers are capable of, and not something that will deliver benefits in isolation.
Achieving the benefits outlined will rely on the union of AI and a distinctly human element. The ability of these algorithms to accurately predict and model chemical entities and related mechanisms rely on our own understanding. This requires a joint effort of both computational scientists and traditionally trained chemists with the necessary understanding of AI and small molecule synthetic mechanisms, respectively.
Additionally, even with a perfect algorithm, existing processes and analytical methods will need to be augmented by AI rather than replaced by it. Testing the predictions, models, and analyses generated by these algorithms necessitates real-world experimentation. This is especially critical when it comes to essential wet chemistry processes, in vivo animal model testing, and heavily regulated clinical trials that are too complex to be accurately replicated by AI.
Looking to the future
What is possible in new small molecule drug discovery, development, and manufacturing is continuing to expand with advancing computing power and a spectrum of possible approaches for the application of AI. Offering a wide array of benefits, including increased speed, lowered costs, and minimized risk, it can be expected that AI will become an integral part of drug production at all steps. However, achieving the full advantages will depend on our own capabilities, willingness, and ingenuity – how we foster these algorithms and how we test and expand their value.
Petra Dieterich, Scientific Lead; Ian Glassford, Director Project Management; Francisco Velazquez, Director, Chemistry.
- Mullard A. The drug-maker’s guide to the galaxy. Nature. 2017;549(7673):445-447.
- Hemmerich J, Ecker GF. In silico toxicology: From structure–activity relationships towards deep learning and adverse outcome pathways. WIREs Comput Mol Sci. 2020;10:c1475
- Coley CW, Thomas DA 3rd, Lummiss JAM, et al. A robotic platform for flow synthesis of organic compounds informed by AI planning. Science. 2019;365(6453):eaax1566.
- Rantanen J., Khinast J. The future of pharmaceutical manufacturing sciences. J. Pharm. Sci. 2015;104:3612–3638.