Taking The First Steps Towards Digitalizing Biopharma Development
By Tyler Menichiello, Chief Editor, Bioprocess Online
Like humans, artificial intelligence (AI) is not perfect, and there are plenty of valid concerns around its use across sectors. However, there is no denying its current and growing potential to improve the way biopharmaceuticals are developed and manufactured. And while wide-scale adoption is slow in this highly regulated industry, companies like AbbVie are forging the path forward and setting an example of what’s possible when these advanced technologies are integrated into development.
In the Bioprocess Online Live event, “How Digital Tools Are Accelerating Biopharma Development,” AbbVie’s head of biologics purification development and digitization teams, Moiz Diwan, Ph.D., explained the company’s lab-in-a-loop model and shared how the company is implementing automation, digitization, and modeling to improve process performance, achieve greater production scale, and progress towards the use of true digital twins in product development.
Setting Priorities Ahead Of Digitalization Initiatives
Both Diwan and Kozyrytska agree that digitalization begins with two main objectives:
- Choose where to start and what to prioritize: According to Diwan, it’s important to assess and understand where the company sits and outline what it specifically hopes to achieve. Trying to do everything at once is impractical and, ultimately, will not be helpful, he said. Set priorities and let them guide your organization’s early decisions on what to buy, where it fits, and what the clear ROI will be.
- Build the financial and cultural foundation: Across the industry, stakeholders are pushing C-suites towards AI use, and that pressure can be leveraged to execute on these digitalization efforts, said Kozyrytska. This excitement at the top allows budgets to be built around the adoption of these technologies, but without everyone on the same page, execution will be impeded. It is for this reason that addressing workforce resistance is important, both Diwan and Kozyrytska argued, as shifting an organization’s culture to embrace the change of digitalization will help smooth the path of execution. Setting a culture to adopt new technologies with the right priorities in mind is equally as important as the technologies themselves, said Diwan.
“I think oftentimes, our scientific organization may not even know how we can digitize our workflows,” said Diwan. This collaboration helped illustrate the scientists’ needs and use cases for the IT team so they could identify where technology would provide the greatest benefit.
Good Data Management Is Foundational
In Biopharma, data is gold. It’s the lifeblood of any pipeline and process, and without good, reliable data, no AI model can be expected to provide value. Security is a crucial part of ensuring data integrity, and as such, should be a cornerstone of any data management strategy, according to Kozyrytska.
“I would lean into the security and morality of the algorithms that we’re developing and implementing,” she said. The data that feeds these algorithms will impact outcomes, so it’s important to feed them data that you trust.
“Curation of that data is really key,” Kozyrytska explained. She suggested organizations prioritize on-premises systems as an additional safety feature to maintain uncontaminated data sets and algorithms in-house.
Next to security, traceability is a key component of good data management. According to Diwan, regulatory agencies are becoming more concerned with traceability, looking for data that tracks back to specific instruments and even individual electronic laboratory notebooks (ELNs).
“That’s something that is becoming more and more critical for us as an organization,” Diwan said about traceability. Ensuring data integrity across systems and having the correct sources identified in regulatory filings has become a key part of AbbVie’s strategy, he explained. Capturing quality data, structuring it, and maintaining its integrity are “the foundational pieces” before building out more advanced technologies like AI and digital tools, Diwan told the audience.
Building For Flexibility And Compatibility
Once the foundation is set for your company’s digitalization initiatives, it’s important to execute with both flexibility and compatibility in mind. Evaluating your company’s current systems and how new technologies will fit into them is critical, according to Diwan.
“You cannot make a change on every single tool,” he explained. They need to be brought in over time, and in some cases, be custom-built around certain off-the-shelf systems. “Having that ecosystem of flexibility is important for us,” Diwan said.
Answers To Your Questions
During every Live event, we encourage the audience to submit questions to our experts via the Q&A feature. Unfortunately, we don’t always get around to answering everyone. However, Diwan and Kozyrytska were kind enough to write responses to several unanswered questions.
These audience questions and our panelists’ answers below have been edited for clarity.
Where do you think digitalization has been more influential, in discovery or in process development?
Kozyrytska: There has been disproportionately more investment in AI for discovery compared to process development. We have yet to truly see the impact of de novo assets in the clinic. When it comes to process development, the impact can arguably be greater due to the nature of processes and the opportunities to standardize. However, the debate is ongoing as to the best way to model processes (mechanistic vs. empirical vs. hybrid). Likely, with more data across molecules, technologies, and processes, the answer will emerge. Big Pharma companies, which have the most data, are actively investigating this question. It is likely that collaborative initiatives, such as the National Institute of Standards and Technology (NIST), may help answer these questions.
Diwan: There is some early work at AbbVie where we are leveraging structure–property relationships to determine biologics’ manufacturability using structure and sequence information, physics, and AI-based models. We have successfully implemented this workflow in certain downstream purification areas, especially in chromatography unit operations.
What are the digital tools AbbVie is using to accelerate process development?
Diwan: We have been implementing various tools to accelerate biologics process development, from mechanistic modeling (e.g., metabolite engineering) to computational fluid dynamics, chromatography and membrane modeling, hybrid modeling in cell culture, and Bayesian optimization tools for more efficient process development and process characterization experiments.
As mentioned earlier, we have some good PoC now in AI-based modeling to relate structural properties to manufacturability. Additionally, we have various digitalization tools to capture, centralize, and visualize our data, which feeds into a modeling tool.
In the commercial environment, we are leveraging various custom and off-the-shelf tools to track and trend manufacturing data, with a vision to build model-based controls in both upstream and downstream unit operations. We believe automation, modeling, and digitalization need to be considered in an integrated manner and developed as end-to-end workflows to maximize the benefits of these enabling tools to truly accelerate process development.
Why are AI tools in biopharma generally geared towards drug discovery and clinical trials, but not so much towards drug development?
Kozyrytska: Indeed, it would be tremendous to see greater investor support for AI tools in drug development. The appeal of AI-powered tools for drug development is three-fold: First, there is immense value in accelerating drug development. Speeding up development not only brings relief to patients sooner, but it also delivers revenue faster and extends a molecule’s effective patent life. Second, the risk profile is better for AI technologies focused on optimization rather than de novo design. Third, the differentiated power of AI is in pattern recognition, in seeing beyond what reductionist-physics-based approaches can offer. The opportunity for this novel type of insight in development is vast. Given all of this, I hope to see some exciting funding rounds come through this year.
Diwan: AbbVie is making an exception by leveraging these digital and AI-based tools to accelerate process development. There are challenges to fully leveraging AI tools due to higher regulatory oversight and the need to ensure that data is captured and analyzed with FAIR principles (Findable, Accessible, Interoperable, Reusable). The other challenge is that, historically, lab data is not structured or formatted in a way to use for developing any AI tools. This highlights the need for companies to focus on data strategy before building their AI strategy.
There is an abundance of AI startup companies popping up. When it comes to collaborating with these companies for data digitalization, is there a process?
Kozyrytska: I agree that there needs to be a standard approach to digitalization and the application of AI. Therapeutics developers need to be equipped to ask technology companies questions about data usage, model development, and IP. As it stands today, there is a shift in IP ownership towards technology companies for which biopharma may not be prepared, and which is likely to impact the M&A of small therapy developers. This is a prime opportunity for a cross-sector collaboration to establish best practices.
Should IT and digitalization teams report directly to QA to make the implementation of systems better streamlined to regulatory expectations?
Kozyrytska: Reporting structure is only one lever in an implementation initiative. In a matrixed organization, both the direct and the indirect reporting lines have to be fully operational, especially during change management. Establishing formal processes for continuous communication and working at the people level to drive adoption of these processes is the more important aspect.
Diwan: There has to be a hybrid model to maximize the collaboration and engagement between IT, process development, manufacturing, and quality. At AbbVie, we have embedded a model where IT sits with technical teams to understand their needs and better develop solutions that are amenable to deploy and leverage sustainably. At the same time, we have hired senior digital strategy leads in technical organizations to work collaboratively with IT teams to ensure that solutions are built with the right rigor and robustness in mind and are tailored to the team’s needs. This hybrid model has been working very successfully at AbbVie.