Supporting regulatory compliance with AI

Ahead of the Paperless Lab Academy event, keynote speaker Toni Manzano shares his thoughts on the use of AI in pharmaceutical manufacturing.

Can you tell me about yourself and your experience in pharmaceutical manufacturing?

I am the co-founder of Aizon. My partners and I created the company eight years ago in Silicon Valley. We work precisely with just one aim. Improve pharmaceutical industry processes using modern technologies, including big data and AI capabilities. We want to enable scientists to be sure that every single drug is delivered at the right time, for the right patient.

In the pharmaceutical industry, there are a lot of inefficiencies. And the reason is that there is not an urgency to change. The pharmaceutical industry earns a lot of money every year, so there is not any kind of urgency to change the existing process. But the pharmaceutical industry, believe me, is very inefficient. Ultimately, these inefficiencies are passed along with the price of the drugs. The final users, in this case, patients, are paying for these inefficiencies.

Eight years ago, we created Aizon to change this mindset, increase knowledge, and optimise processes in pharmaceutical manufacturing to ensure that every batch created is the best. We knew that we could only do that using AI.

Is the Pharmaceutical industry slow to adopt change?

I wrote an article one year ago where my colleagues and I described and calculated the gap between the pharmaceutical industry adopting innovation in regards to other industries. The result of this paper was that the pharmaceutical industry, on average, is 11 years behind the other industries when adopting innovation. Sometimes you might see the pharmaceutical industry pointing to the regulator as the root cause of a lack of innovation. The shocking result was that regulators are faster than the industry to adopt innovation.

We're spreading the word. We are trying to teach people at conferences, publishing articles and so on in order to democratise access to this knowledge. Nevertheless, I would like to highlight something that probably it's important, in comparison with the other industrial revolutions; the difference with this one is society is driving this new change.

Society is pushing the industry to innovate. In the previous industrial revolutions, it was the industry that was driving innovation. Society adopted innovations that were introduced by the industry. But in the case of this fourth industrial revolution, it is society that is using 5G, it is society that is using AI everywhere. Starting with the beginning with the smartphone, today we are using augmented reality, virtual reality, big data, and cloud technologies. Every day society is using and emphasising the need for these technologies. And I firmly believe that industry, in this case, the pharma industry, will adopt AI because society is driving that adoption.

Today pharma organisations are using Microsoft Teams, Google Meet, Zoom, or shared drives, and all these things are in cloud. So they are adopting that because the rest of society is pushing.

OEMs that provide equipment such as bioreactors, filling lines, and packaging lines are already providing the equipment, with sensors connected or the capability to connect data to the cloud. Many opportunities are provided to laboratories: the ability to control the bioreactors to detect potential anomalies before they happen; to predict anomalous behaviours in the industrial reaction. As these kinds of opportunities increase in number, eventually, the pharmaceutical industry will be convinced that they have to change.

How can AI improve GXP compliance?

How to improve the GXP in general? In the end, if we can automate the control of all the processes, scientists and lab managers can be sure that everything is performed as expected.

We are improving this inefficiency, and we are improving the quality behind the GXP requirements. AI does not replace humans. It enhances human capabilities and allows humans to be dedicated to delivering higher value than just analysing things that machines can do better.

Inspection is another example. Nowadays, AI systems interact with filling lines or vials to detect anomalies in the packaging vials. This is another way to improve the quality.

Lab managers consider manufacturing in pharma operations as something that is always constant. This is not real. The raw material can vary depending on the provider; the processes depend a lot on the instrument and interaction, the setup with operators and so on. Everything is changing. So why not consider the variations, the variability in the process? AI brings a lot of opportunities to improve GXP conditions.

Humans interact with a reactor based on recipes. But these recipes may not consider all the potential variabilities on raw material, sugar, the material that we are using to deliver insulin or many other raw materials that we are using to acquire this insulin or, for example, lipids, proteins and vaccines.

When humans interact with these static recipes, they try to accommodate what happens during a 15-day bio-reaction as best they can with their own knowledge.

But the process is long, there are different shifts, and there are a lot of human interactions depending on the know-how of each person interacting with the process. These inefficiencies come when we have different behaviours, attitudes, or interactions with batches, based on people's experience with a process.

How is compliance changing over time?

In March, the US Federal Drug Administration (FDA), proposed a draft to implement AI in drug manufacturing. They aim to evaluate the technology and to try and introduce it in a very regulated way. The FDA is moving ahead to promote a technology that the pharma industry is still considering.

This is just one example, but I can give you another. Two years ago, the FDA published the guidelines to implement AI in medical devices, not in manufacturing, but in medical devices. And they were considering a lot of different topics. The FDA is pushing in this direction, to use AI to have more knowledge, more knowledgeable actions, and improve scientific research and manufacturing.

The United Nations has different workshops to teach the pharmaceutical industry how to use AI in the right way, using good practices. The regulatory bodies are ready to help. The problem is that they can only do something if the pharmaceutical industry adopts the technology in an industrial context.

The regulatory bodies are ready, the administration is ready, and the compliance is ready. The problem is that the industry doesn't want to move forward.

Toni Manzano is the co-founder and Chief Scientific Officer at Aizon. The company has designed a cloud platform that provides simple solutions for analysing and improving industrial processes in Biotech and Pharma manufacturing.