Olivia Friett spoke to Aaron Johnson, manager, cheminformatics & data science at Lonza about the role of AI and ML in drug development.
Lonza
1. How are artificial intelligence (AI) and machine learning (ML) being used in drug development and manufacturing today?
The pharmaceutical industry is undergoing a transformative shift driven by advancements in artificial intelligence (AI) and machine learning (ML). These technologies are addressing the increasing demand for faster and more efficient therapy development, ultimately reducing costs and timelines for customers.
In drug discovery, AI algorithms analyse vast datasets of biological and chemical information to identify potential drug targets and predict the properties of candidate molecules. This accelerates the traditional drug discovery process, enabling researchers to focus on the most promising leads and reduce the time and cost associated with experimental testing.
In manufacturing processes, AI and ML play a crucial role in optimising production processes, ensuring quality control, and predicting equipment maintenance needs. By analysing real-time data from manufacturing lines, AI can identify and correct process deviations, maximising yield and minimising waste. Predictive maintenance algorithms anticipate equipment failures, reducing downtime and ensuring a stable supply chain.
Overall, the integration of AI and ML not only improves the efficiency and cost-effectiveness of drug manufacturing but also enhances product quality and safety, leading to better patient outcomes.
2. What role does AI play in route design and overall timelines for route scouting?
AI significantly accelerates and refines route design and scouting in drug development, which is traditionally a time-consuming process. AI algorithms can quickly analyse vast amounts of data, enabling drug developers to explore a wider range of synthetic routes, identify potential bottlenecks, and optimise reaction conditions in silico before committing to costly lab experiments. Furthermore, combined with existing information, AI can be used to predict more commercially viable starting materials early in the route design process.
In terms of timelines, AI significantly shortens route scouting by minimising the number of synthetic steps required to produce drug candidates. By automating the analysis of complex chemical data and predicting reaction outcomes, AI reduces the need for extensive trial-and-error experiments. This allows chemists to quickly act on the most promising synthetic routes, leading to faster development timelines and reduced costs. AI can also assist in generating alternative synthetic pathways, allowing for contingency planning and mitigating risks associated with supply chain disruptions or unexpected reaction outcomes.
3. Are there any challenges in using AI and ML-powered tools?
Despite the benefits of using AI and ML-powered tools for drug development, barriers remain.
One of the key challenges is the availability of suitable data. Often, the accessible data may be limited, of low quality, or inconsistent, which can affect the accuracy and reliability of the results. Additionally, there are concerns about fairness and bias. If the data used to train an ML algorithm are biased or unrepresentative, the resulting predictions may be inaccurate or unfair.
It is essential for drug developers and their partners to ensure the availability of high-quality data for more comprehensive models and to implement appropriate guardrails for data protection and collection.
In addition, there is a need for interdisciplinary collaboration. Effective deployment of AI and ML in drug development requires chemists, biologists, data scientists, and engineers to work together seamlessly. Bridging the gap between these diverse fields can be difficult, and fostering a collaborative environment is essential for maximising the benefits of AI and ML technologies.
4. How is Lonza utilising AI and ML to streamline the drug development and manufacturing process?
At Lonza, we're actively integrating AI and ML throughout the drug development and manufacturing process. For instance, we're implementing ML algorithms and AI to navigate the complexity and speed requirements of manufacturing novel treatments and accelerate the synthesis of APIs. One example is our AI-enabled route scouting offering, which combines extensive, proprietary commercial data in the industry-leading computer-aided synthesis planning technology.
The new Route Scouting Service offering combines Lonza's process R&D expertise and supply chain databases with technology tools from multiple partners. With access to global chemical supply chain intel and the predictive power of award-winning AI, the new offering provides synthetic pathways that are more supply chain resilient and offer insights for optimal route design on both clinical and commercial manufacturing. The integrated service can provide intellectual property, reduce COGS and improve supply chain security for customers.
Additionally, in small molecule development and manufacturing, ML is used for synthetic route optimisation, retrosynthesis, toxicological assessment of new chemical entities, and formulation design. In drug product development and manufacture, ML is employed in developing controlled-release tablets to assess the hardness, particle size, moisture, and other factors that predict a tablet’s in vitro behaviour.
5. How do you see the role of AI and ML evolving in the drug development and manufacturing space over the next five years?
As AI and cheminformatics capabilities continue to advance, several developments have the potential to further revolutionise drug development and manufacturing processes. For example, predictive modelling could be expanded to optimise synthesis routes not just for efficiency and cost but for plant fit, utilising existing assets and equipment to the fullest. Additionally, AI-driven retrosynthetic analysis could integrate real-time production data to dynamically adjust routes based on current plant capacities, inventories, and operational constraints.
When looking at ML, models trained on vast datasets of reaction data could enable de novo design of novel synthetic routes and methodologies, creating new synthetic possibilities. Cheminformatics tools may provide deeper insights into impurity prediction and formation, enabling proactive mitigation strategies. Ultimately, the integration of AI and ML could result in processes that can dynamically adapt and self-correct in response to disruptions, all while identifying the ideal pathway from both a chemical and operational standpoint.