AI Stuck Upstream? It Still Influences Manufacturing Outsourcing

By Louis Garguilo, Chief Editor, Outsourced Pharma

Artificial intelligence may be helping to run and optimize bioreactors at your CDMO. Certainly, there is pervasive pressure to move manual (and reactive) control to AI-autonomous (predictive systems).

You may already select your CDMO partially based on its systematic utilization of artificial intelligence applied to its manufacturing facilities.

Perhaps you peer at your digital twin through an AI looking-glass to monitor conditions, adjust parameters … and realities.

Perhaps.

But at this, our moment of transition, the odds are that AI still sits on the doorstep. It is not quite securely ensconced within your CDMO, actively figuring out why your yield just dropped (or maybe increased) by 20%.

Yet while AI may not fully reside in the manufacturing plant today, it increasingly shapes the decisions determining what and how programs get to and proceed through that plant.

That’s the through line from a recent conversation with Jon Williams, CEO, Lumanity, an organization focused on "value demonstration" across the drug development and commercialization life cycle. Lumanity combines human expertise with “advanced technology and AI-based tools to provide end-to-end commercialization support.”

Let's see what that might mean.

The Industry’s Broken Equation

Williams starts with our age-old frustration.

 "It still takes between $2 to $3 billion to bring an asset to market over many years,” he intones. At the same time, given the nature of our advanced medicines and the varying number of addressable patient markets, “our understanding of medicine has become more precise.”

Despite those markets and that understanding, budgets haven’t budged. Costs are still high, timelines still long.   

Our first life raft – “big data” – did not rescue us as we had hoped. AI alone is not going to be a savior either, says Williams.

What will save us, he says, “is a trifecta of higher quality data, technologies to unlock insights out of that data, and human experts asking the right questions and making better decisions.”

Data, technology, human expertise. A nice combo, until now practiced mostly at the front end in the drug and therapy life cycle.

But Williams believes bending the time/money curve there will affect the candidate trajectory downstream through development, and our manufacturing ecosystem.

And as we all know, that latter ecosystem is reliant on external partners.  

Where AI Shows Up

At Lumanity, AI is not necessarily applied directly to manufacturing, but guides that down-stream decision-making by:

• Synthesizing available real-world evidence
• Accelerating literature reviews
• Generating economic (market/patient) insights
• Improving how evidence and insights are translated into regulatory and scientific communications

It’s about turning massive datasets into what Williams calls “rapid, transparent, accurate, actionable intelligence" that feeds discovery through commercial. 

To accomplish this, Lumanity has created an AI system that selects the best available AI model for each use case.

“We don’t perceive our value-add is creating the ten-thousand-and-first AI model. It’s the ability to choose the best of those models,” he explains.

Sounds great, and rather research- and discovery-based.

However, Williams explains, if AI improves how we define patient populations, predict demand, and demonstrate financial value, it inevitably influences how we manufacture, how much commercial product is needed, and at what pace of that production.

Those improved analyses should thus flow into CDMO identification, contracting, and ongoing performance metrics. Your “early” AI model encompasses your later external manufacturing, and CDMO selection.

To be clear, Lumanity does not target the goings-on at a CDMO’s pilot plant or API facility. But, says Williams,  “You can’t start to think about a commercialization strategy without thinking about the manufacturing and development strategy.”

Makes sense; with advanced modalities like cell and gene therapy, material creation is the business model.

CDMOs for their part have been trying to reach upstream to engage customers earlier. CDMOs today advise on development strategy, and want to help interpret data and guide process design for customers.

And increasingly, CDMOs are deploying their own digital tools to do so.

Therefore, as touched on above, soon you may differentiate prospective partners based on their AI agilities.

“If my CDMO doesn’t have this capacity,” says Williams of the coming sponsor attitude, “we’re going to the one that does.”

Certifying Trust For AI Adoption

The distance, shall we say, between AI and trust may be shortening, but in our highly regulated environment, putting the two together is remains a challenge.

Williams says that has been exacerbated by “a lack of transparency and a lack of auditability” of AI tools.

To address that, Lumanity obtained ISO 42001 certification, the first company in the space to do so. Williams believes the time is coming when this may be required at all one’s outsourcing partners.

That’s an interesting proposition. If we think about it, we already demand strict cGMP compliance and full traceability – we demand that trust factor. Why would our demands on AI-enabled decision systems be any different?

Williams says the sponsor and CDMO that can demonstrate validated, auditable AI integration will have a competitive advantage over others less able to do so.

I wonder whether at some point the FDA and other regulatory bodies themselves require AI certifications such as ISO 42001 at facilities.  

Maybe the better question is why wouldn’t they?

“Big Data” To Real Intelligence

There's both a profusion of nuance and broad strokes of conjecture applied to AI, and so, says Williams, inevitably we are all striving for “actionable intelligence.”

He's talking about a progression starting with, and continuously generating and collecting data; analyzing and understanding it (the human element); and then acting on it with full confidence in the entire chain of those events.

That chain of confidence must flow to and from external partner, feed development and processing, and enlighten and embolden product and business strategy.

Williams tells me he established Lumanity to “better define and demonstrate value in the entire process of drug and therapy commercialization.”

Such analysis – even when arrived at early in the cycle – will influence a biotech's entire business plan, and your CDMO selection and outsourcing operations.