The biopharmaceutical industry is no stranger to artificial intelligence (AI) and machine learning, particularly in the discovery lab. But AI could also improve manufacturing, said Julia Rozenbaum, business development researcher and data analyst at Sigmoidal, a consulting group.
“AI has the potential to impact each step of the drug production process. It might be the most crucial to the first stages,” she explained. “Algorithms can predict interactions between molecules to understand mechanisms of disease. This can help to find new drug components. AI might also be of use in running preclinical tests. That could shorten the time to develop and launch new medicines.”
AI also has the potential to streamline processes on the biomanufacturing facility factory floor and distribution center, said Rozenbaum, noting that “When it comes to the manufacturing process itself, AI can process in real-time the vast amount of data that is being produced. It can lead to improved supply chain planning, forecasting, inventory management, and basically help to speed up and automate every step of the process.”
Barry Heavey, a managing director in Accenture’s life sciences practice, also sees a role for AI in manufacturing.
“In the future, AI could help deconvolute the complex biology in the bioreactor—helping to understand the correlation between the culture conditions, the physiology of the producer cells, and the yield and quality attributes of the final product,” he told GEN. “AI will also help manufacturers gain a better understanding of causal relationships, which in turn may help develop more consistent, higher yield processes more quickly in process development and speed up root cause determination when there are issues with existing processes in routine commercial manufacturing.”
But the biopharmaceutical industry is yet to embrace AI. In fact, the approach is not even close to being commonplace, according to Heavey, who said the industry is struggling to find the best way to record, manage, and interpret the vast amount of data generated during the manufacturing process.
“We are still in early days when it comes to using AI in biotech manufacturing. Indeed, some companies are still trying to roll out standard multivariate analysis” he added.
Biopharma’s AI challenges
According to Daniel Faggella, founder of Emerj Artificial Intelligence Research, biopharma struggles to implement AI on the factory floor.
“As far as I know, no company has integrated AI so thoroughly as to have either massively improved, or massively screwed up their operations. Of course, that information would be hard to find either way,” he pointed out. “There are a lot of initiatives and investments, but little by way of concrete steps. One of the major challenges is determining how to integrate AI into workflows. Which phases of the process can we use AI to help improve human decisions? It’s all nascent so it’s hard to pick. It’s opportunity but also risk.”
Another difficulty is the lack of overlap between biopharma and computer science, Faggella said.
“Computer science and life science don’t communicate well,” he continued. There has not been enough ‘osmosis’ to allow these two groups to feel part of the same team. They are a foreign language, fighting for another team. Integrating cross-functional teams to work on pharma data science problems is critical.”
Smaller, innovative companies will lead the way in AI, predicts Faggella.
“I’m of the belief that we may need to build the pharma firm of the future from the ground up, like Benevolent.AI is trying to do—because the AI deployment in these behemoth firms is so slow and painstaking.”