Advancing the pharmaceutical supply chain using artificial intelligence

Share this article

Two people looking at a digital supply chain simulation
Rapidly spreading diseases and unpredictable global shifts cause increased strain to the pharmaceutical supply chain, jeopardizing the secure delivery of drugs to those who are in need. Artificial intelligence can maximize efficacy of the value chain and allow optimal responses to changing market demands. Two experts share insights into the possibilities and hurdles associated with integration of machine learning tools into the pharmaceutical supply chain.

Integrating multidisciplinary links into one chain

After identifying and optimizing an active drug substance for clinical efficacy and safety, several additional steps are still necessary before a company is able to produce a marketable drug product. The supply chain encompasses all activities involved in delivering an end product to the patient. It integrates factors such as product stability, traceability, storage capacity, and flexibility in response to unforeseen events. In the occurrence of an unexpected disease outbreak in a particular country, for example, the supply chain of pharmaceutical companies must manage the subsequent surge in drug demand.

To deliver the appropriate dosage, in the proper location, and at the appropriate time, it is vital to have a strong interaction between the supply chain and other components of the drug development process. For example, a demand for increased drug stability expressed by clinical trial operations or a logistics department can be accomplished through a collaboration with pre-clinical development units where the manufacturing or formulation strategy is to be optimized. The supply chain functions as a complex hub of interactions, both internal and external, which is fundamental in linking knowledge to outcomes.

Enhancing efficiency and sustainability

Artificial intelligence (AI) and machine learning (ML) present a significant opportunity to enhance the effectiveness of the supply chain. The time has come to explore their potential to the fullest. Christian Baber, Head of Scientific & Pharmaceutical Data, Informatics & Systems at The Janssen Pharmaceutical Companies of Johnson & Johnson, agrees: “Both Covid-19 and the war in Ukraine have put supply chains to the test. AI is great at improving robustness and resilience – it allows us to react more quickly to unforeseen challenges, ensuring uninterrupted medication delivery to patients in need.”

Predictive modeling can be used to improve the yield, speed, and flexibility of a supply chain. It allows one to anticipate the response of a system in specific situations. Digital twins are increasingly being used in this context. “The virtual representation of a physical process allows fast and efficient optimization through simulation,” says Baber. By generating algorithms that replicate the behavior of the supply chain, we can use insights to improve processes based on historical and sensory data, and adapt the pipeline based on demand forecasts.

“To deliver the appropriate dosage, in the proper location, and at the appropriate time, it is vital to have a strong interaction between the supply chain and other components of the drug development process.”

Christoph Portier, Manager at CESPE (Centre of Excellence in Sustainable Pharmaceutical Engineering & Manufacturing), sees the potential of data science in numerous stages of the production pipeline. “In the past, operators mainly studied the impact of a single parameter, such as temperature or mixing speed, on the production process of chemical drug substances. Now, we adopt a multivariate approach that utilizes AI to consolidate vast amounts of experimental and sensory data, and provide a more realistic projection of the drug’s behavior.” But we can also optimize on the level of the drug product itself. “At CESPE, we have a unique and extensive library of raw material characterization data. Our algorithms use these data to predict how a specific drug composition will behave in a manufacturing unit operation under defined circumstances,” Portier adds. AI-ML allows optimization of the production process, but also enables timely intervention to prevent significant issues, resulting in an uninterrupted pipeline. “AI tools are increasingly being used for automated quality control, allowing for early detection of problems and concomitant root cause analysis leading to efficient establishment of preventative actions.”

AI-ML tools can also render the supply chain more sustainable. According to the recent Corporate Sustainability Due Diligence Directive proposed by the European Commission, companies will soon be required to demonstrate their concrete efforts in safeguarding the environment. “With the help of AI- ML, companies can assess and optimize factors like CO2 emissions and power consumption to meet these requirements,” Portier states.

Towards pre-competitive collaboration

One of the main challenges when implementing AI-ML in the supply chain is the level of complexity. “Although using a digital twin in the production process comes with great advantages, it is still hard to accomplish. The more complex the physical properties under study, the harder the optimization gets,” says Portier. “Adoption also needs to happen in real time,” adds Baber. “Data needs to be processed at a considerable speed for it be relevant in process optimization.” It is, moreover, not always easy to validate the strategy used, as the algorithms are complex and hard to interpret.

Another obstacle is the scarcity of high-quality data. “If you rely solely on your own data to generate a model, it can take years before you collect sufficient amounts for relevant model building,” states Baber. “Alternatively, you can buy data, but this increases the implementation cost of AI-ML.” A more democratized and cost-effective strategy is to collaborate with partners in a consortium, although there are challenges associated with sharing data as well. “A collaboration between a company and an academic institution is often possible, but it becomes considerably harder when multiple companies active in the same sector are involved,” states Portier.

Read this article to learn more about unlocking the potential of patient data using healthcare data standards!

Both Baber and Portier agree that pre-competitive collaborations are a great way to work around this IP issue. In these projects, multiple companies share knowledge that is either generic or off-patent, without directly competing with each other. As true innovation lies in bringing together expertise, it is necessary to foster collaborations. “It is crucial for funding agencies and regulators to have an understanding of the needs within the ecosystem and actively engage in discussions regarding the allocation of funds,” states Portier. “The dedicated calls of both MEDVIA and BioWin, designed to enhance the adoption of innovative technologies within our Belgian ecosystem, stem from their active interaction with all stakeholders.”