The biopharma industry is embracing artificial intelligence (AI) in manufacturing, even though questions remain about how best to use the technology in a GMP environment.
At least, so says Sanjay Konagurthu, PhD, senior director, science and innovation, pharma services, at Thermo Fisher Scientific, who argues that the key to successful AI adoption is a clear use case.
“As is true in most industries, adoption of AI and machine learning (ML) is real and accelerating across the biopharma industry. However, most use cases center around applications that augment teams without completely redefining a validated process, such as smarter quality and inspection workflows.
“We’re seeing that the hesitation isn’t so much reluctance to adopt AI as it is the practical constraints of operating in a good manufacturing practices-regulated environment. To adopt AI and ML in regulated environments, you need clear intended use, strong data foundations, traceable governance, and set parameters for disciplined control and monitoring,” he tells GEN.
Data foundations
In addition to establishing a strong use case, drug companies need an IT infrastructure that facilitates the flow of process data, according to Konagurthu, who says data silos are a persistent problem in biopharma.
“As with most scientific endeavors, vast quantities of data are generated across the biopharma industry, among labs, organizations, consortia, and nations, and much of this data is stored in a singular system. So, there’s a connectivity challenge, but that’s not the only reason why processing technologies struggle to exchange information.
“The data is often captured according to different standards and exists in a variety of formats, which means context is easily lost. Also, the data exists in a wide variety of structured and unstructured formats, which compounds the challenge in effective curation and analysis,” he says.
Failure to establish an effective infrastructure or standardize data has multiple negative consequences, Konagurthu adds.
“When data from early development can’t be connected through to commercialization, teams end up re-running experiments and analysis. They may even miss early signals that could impact downstream manufacturing or risk quality.
“When companies look to scale or implement new technologies like AI and ML, fragmented data can become prohibitive. Ultimately, for biopharma, this could extend the time it takes to bring a promising molecule to market,” he says.
Formul-AI-tion development
Beyond process development and control, formulation is another area where more and more biopharmaceutical companies are making a use case for AI.
Konagurthu says, “Biopharma scientists have historically used trial-and-error approaches to determine the right solubility and bioavailability of OSD [oral solid dose] therapies. With AI and ML models, teams can make earlier, better-informed decisions on formulation pathways.
“Early-stage acceleration in discovery and formulation echoes all the way into manufacturing and clinical supply, so improving the front end can compress timelines across the entire pipeline,” he adds.
