What hidden engine drives biopharma's relentless pursuit of faster, more reliable, and cost-effective therapies? While many innovations grab headlines, at the heart of this transformation, particularly in biologics manufacturing, lies the often-underestimated yet profoundly critical area of AI-driven cell culture media development.
For decades, the composition and optimization of culture media have remained closely guarded secrets, a testament to their pivotal role in defining the success of a bioprocess. This fundamental appreciation for culture media has grown significantly over the past 30 years, shifting from a mere purchased commodity to a recognized strategic component.
A leading expert in the field, Tom Fletcher, Research and Development Director at Fujifilm Irvine Scientific, offers invaluable insights into the fundamental shifts in biomanufacturing, current challenges in cell culture media development, and, critically, how automation, machine learning, and artificial intelligence (AI) are reshaping the industry.
The overarching theme highlights a crucial balance: how can biotech scientists effectively leverage these cutting-edge technologies while ensuring they enhance, rather than replace, human understanding and expertise in bioprocess development? This is not merely a question of adopting new tools but of fostering a new paradigm of collaborative intelligence.
This concept is discussed in greater detail in an episode of the Smart Biotech Scientist Podcast, hosted by David Brühlmann, founder of Brühlmann Consulting.
The Evolving World of Cell Culture Media Development
Tom Fletcher's career has paralleled, and indeed contributed to, the dramatic evolution of cell culture media. Thirty years ago, the focus was primarily on serum-free media, a significant advancement at the time. Yet, as Fletcher explains, the industry quickly progressed, driven by demands for greater safety and regulatory compliance.
Customers began to request the elimination of animal components, prompting developers to innovate with alternatives such as vegetable-derived protein hydrolysates. This relentless drive for purity and definition has culminated in an era where chemically defined media are now the norm, representing a monumental breakthrough. As Fletcher points out, it's difficult to envision a "next category" beyond chemically defined, as the components are fully known.
However, even in this highly defined environment, challenges persist. Modern media development isn't just about identifying what's required for the process; it's about mastering the intricate interactions between 60, 70, or more components, especially as processes intensify. Higher cell densities and increased concentrations, particularly in feed media, introduce complexities like redox reactions.
These reactions, often slow and unanticipated, can lead to unexpected issues, such as the filtration of critical metals. The importance of controlling impurities, even trace metals like manganese, has also become increasingly apparent, as they can significantly impact product quality, such as glycosylation.
Moreover, the practicalities of media preparation, though seemingly primitive, remain significant hurdles. Learning to concentrate complex solutions, particularly feed media, and developing stabilization methods for these concentrated media (crucial for perfusion processes) are ongoing challenges. The historical secrecy surrounding media formulations, which Fletcher views as a compliment to the field's importance, underscores the proprietary knowledge embedded in these solutions.
Despite these complexities, significant progress has been made in the approach to media development. The industry has shifted towards adopting platform culture media for speed, a practice that was uncommon three decades ago. This allows for rapid entry into the clinic, followed by planned optimization in later stages, demonstrating a mature understanding of the development lifecycle. This combination of platform reliance and strategic optimization reflects a pragmatic approach to accelerating biomanufacturing.
AI in Bioprocessing: Promise, Pitfalls, and the Pursuit of Knowledge
The integration of emerging technologies, such as automation, machine learning, and AI, is profoundly shaping the industry. Tom Fletcher, while acknowledging their tremendous promise, offers a balanced perspective, cautioning against both unwarranted skepticism and excessive reliance. He notes that he has heard from experienced professionals, particularly in late-stage process development, who dismiss these new tools due to the inherent complexity of bioprocesses and the scarcity of proven large-scale success stories.
Conversely, he also observes an overoptimistic view, where some believe that sophisticated in silico algorithms alone can efficiently resolve process development issues, requiring only the "push of the right buttons." Fletcher disagrees with both extremes, asserting that these tools are invaluable when applied correctly.
The core of Fletcher's philosophy on integrating AI lies in the concept of Collaborative Intelligence (CI): Artificial Intelligence (AI) plus Human Intelligence (HI) equals Collaborative Intelligence. This means that AI should never minimize the influence of human experts or neglect the invaluable experience that generates true knowledge. What is produced in R&D, Fletcher emphasizes, is knowledge, not just information or interesting outputs from a machine. While AI can learn and generate data, the leap to true understanding and expertise requires human insight.
A recurring theme is the critical distinction between merely "picking winners" and generating knowledge. Fletcher draws a parallel to the advent of automation and high-throughput methods, which, while successful in quickly screening vast numbers of possibilities, inadvertently fostered a "laziness" among some scientists. Instead of thoughtfully designing experiments to test hypotheses and understand why certain outcomes occurred, the focus shifted to simply getting a quick answer.
"Don't just think of your job as picking winners out of these random experiments; think of your job as generating knowledge. I want to know not which one is best, and I want to know why it's best."
This deep understanding is paramount because it prepares scientists for future challenges, whether it's answering questions from regulatory agencies or troubleshooting unexpected events. The value of this knowledge, though often underestimated in the rush to find quick answers, is immense and lasting.
I think it's interesting that the culture media is one of the most closely held secrets. When you talk to somebody about a bioprocess, they might tell you all about their downstream process, what resins and what they might tell you about their cell line and things like that. But the culture media is like "Oh, no, no, we can't disclose that."
Navigating the "Black Box" and Fostering Collaboration
One of the most significant risks associated with over-reliance on AI is the "black box" problem. AI algorithms often operate in ways that are difficult for humans to interpret or explain, leading to outputs without a clear rationale or justification. This lack of transparency is particularly concerning in bioprocessing, where a thorough understanding of the process is vital for both scientific integrity and regulatory compliance.
Scientists must be able to articulate why a particular decision was made, not just that a machine provided an answer. As Fletcher aptly states, "We're not just trying to get an answer; we're trying to get an understanding of what were the logical steps to arriving at that answer." While a definitive solution for demanding explainability from AI is still evolving, the need for the tool to provide rationale is clear.
The increasing reliance on AI, modeling, and automation also brings about a shift in team dynamics, requiring greater diversity and collaboration. Bioprocessing teams, traditionally comprising biologists and chemists, now include data scientists and experts from varied backgrounds. Fostering effective collaboration among these diverse groups presents a significant cultural challenge.
Fletcher emphasizes that leadership must cultivate a collaborative culture where team members respect and listen to each other, leveraging their combined strengths to achieve common goals. This means being patient, understanding different perspectives, and recognizing that what might seem logical to a bioprocess expert may not be intuitive to a data scientist and vice versa. It's about building bridges of understanding to maximize efficiency and effectiveness.
Strategic AI Implementation: Advice for All Company Sizes
For biotech leaders weighing substantial investments in AI infrastructure, particularly for smaller and mid-sized companies with limited resources, strategic implementation is key. Tom Fletcher acknowledges the difficulty of this challenge, noting the absence of "discounted AI software."
His primary advice for smaller entities is to look outwards and leverage external resources.
This includes:
- Suppliers: Many suppliers are willing to go the extra mile, offering expertise or access to technologies to secure business.
- Consultants: Hiring experts on a consulting basis can provide specialized knowledge without the long-term overhead of a full-time employee.
- Networking: Actively asking specific questions within professional networks can lead to valuable solutions or connections to individuals who can assist.
Entrepreneurs often benefit from asking, "Who can do that for me?" instead of "How can I do it myself?" This mindset allows smaller companies to tap into tremendous potential without having to build every capability in-house. It's about strategically partnering to access the advanced tools and expertise needed for media development and bioprocessing.
AI's Role in Building Resilient Supply Chains
The role of AI in enhancing supply chain resilience is a critical concern, highlighted by recent global disruptions. While concrete examples of AI being directly used in this context may be emerging, its immense power for demand planning is clear. Businesses constantly strive to minimize inventory for financial reasons, but AI, fed with diverse inputs, could transform this.
Fletcher suggests incorporating:
- Sales forecasts: Traditional input for demand planning.
- Global supply factors: Broader geopolitical and economic indicators.
- Historical data: Crucial for identifying trends and patterns.
- Market trends and breaking news: AI could process real-time information to anticipate disruptions.
By diversifying inputs and utilizing sophisticated in silico tools, companies could significantly improve demand planning, better preparing them for unforeseen challenges like a pandemic. Instances exist where companies anticipated disruptions and strategically built up inventory, outcompeting others due to this foresight.
While the direct involvement of AI in those specific instances might vary, its potential to connect market trends and news with supply chain decisions is clear, empowering bolder and more informed executive choices.
Differentiating Transformative Tech from Hype
In an industry prone to technology hype, biotech leaders need a discerning eye to differentiate genuinely transformative AI applications from overhyped areas. Tom Fletcher advises a healthy skepticism when something "sounds too good to be true." His recommendation is straightforward and rooted in scientific rigor: demand case studies and data.
"If somebody's going to recommend or sell you something that sounds amazing, I would first want to see some case studies."
For scientists, the expectation is always to "show me some data." This applies equally to AI. Companies advocating for AI tools should be able to provide concrete, relevant case studies demonstrating their effectiveness. This ties back to the fundamental scientific principle of understanding how something works.
It's not enough for a technology to perform "miracles"; scientists need to look "under the hood" to understand the underlying logic and algorithms. This critical approach ensures that investments are made in solutions that genuinely add value and transparency to bioprocess development.
Embrace AI because I think we all believe it's valuable but don't allow lazy science to creep in. Got to resist it because it'll happen. If you don't resist it, people are going to get lazy and just say "Well, I've got this supercomputer and I know which buttons to push." That's not good enough.
Lessons from a Veteran: Advice for the Future Bioprocess Scientist
Drawing from his extensive experience, Tom Fletcher offers profound advice to those starting a career in bioprocessing. While he might consider studying biochemical engineering today, given the evolution of the field, his core counsel revolves around two timeless principles:
- Embrace Hands-On Experience: There is no substitute for working directly in the lab, where you learn how things truly happen in a bioprocess. This practical, experiential knowledge is invaluable for developing or operating bioprocesses.
- Learn from Failures: Perhaps the most impactful advice, Fletcher emphasizes that failures are where the most profound learning occurs. Difficulties encountered in developing a bioprocess teach lessons that last a lifetime. Instead of dreading failures, one should be grateful for the opportunity to learn from them. These challenging experiences reveal what works and what doesn't, contributing directly to the accumulation of valuable knowledge.
"The failures are where you learn the most. The failures are when you encounter difficulties... You're going to learn some lessons that are going to last you through the rest of your life."
This perspective fosters resilience and a continuous learning mindset, crucial attributes for navigating the complexities and rapid changes in biotech.
The Future of Bioprocessing: An Integrated, Digitalized Approach
Looking ahead five to ten years, Tom Fletcher envisions a future for bioprocesses that, while not "magic," will see significant advancements built on current trends. He anticipates an even greater dominance of robust generic platforms, allowing companies to get therapies into the clinic much faster. This will be complemented by a continued focus on intensification in late-stage development, driven by the increasing importance of cost efficiency.
A significant shift will be the pervasive integration of automation and digitalization. This existing trend will become even more widespread, with processes designed to be automated and data fully digitized. This digitalization will enable crosstalk between unit operations and advanced feedback control systems, leading to more reliable processes. The human role in manufacturing will change, with a lower headcount and a greater need for individuals skilled in managing digitalized and automated equipment. This will require a greater focus on data science, even within manufacturing environments.
In essence, the future of bioprocessing will be characterized by highly integrated, automated, and digitalized workflows, where data flows seamlessly to control and optimize processes. The human contribution will evolve from manual execution to strategic oversight, data interpretation, and advanced problem-solving, underscoring the collaborative intelligence model.
Conclusion
The collective wisdom shared by industry experts boils down to a singular, powerful message: Embrace AI but don't let lazy science creep in.
AI and machine learning offer unprecedented capabilities for accelerating discovery, optimizing processes, and enhancing efficiency in cell culture media development and bioprocessing at a large scale. However, their actual value is unlocked only when coupled with human intelligence, critical thinking, and a steadfast commitment to generating genuine knowledge and understanding.
The temptation to simply "push buttons" and rely solely on an algorithm for answers must be resisted. Instead, biotech scientists and leaders must be challenged to be thoughtful, to formulate hypotheses, to demand explanations from their tools, and to always seek the "why" behind the "what."
By fostering a culture of collaborative intelligence, where human expertise guides AI's power, and by prioritizing fundamental process understanding over quick, unexamined answers, the biotech industry can truly harness the transformative potential of these technologies, ensuring that innovation leads to robust, reliable, and deeply understood bioprocesses for the benefit of all.
About Tom Fletcher
Tom Fletcher is R&D Director, Process development, Fujifilm Irvine Scientific, a company that specializes in the development, manufacture and supply of cell culture media for a variety of applications.
He currently leads the Process Development group, which specializes in tailoring effective solutions to meet the specific needs of Fujifilm clients worldwide. He has over 30 years of experience serving the biopharmaceutical industry, developing cell culture media for large-scale production processes.
Thomas has previous experience in protein chemistry research at the University of California, Irvine School of Medicine, and working at Becton Dickinson & Co.
Connect with Tom Fletcher on LinkedIn.
David Brühlmann is a strategic advisor who helps C-level biotech leaders reduce development and manufacturing costs to make life-saving therapies accessible to more patients worldwide.
He is also a biotech technology innovation coach, technology transfer leader, and host of the Smart Biotech Scientist podcast—the go-to podcast for biotech scientists who want to master biopharma CMC development and biomanufacturing.
Hear It From The Horse’s Mouth
Want to listen to the full interview? Go to Smart Biotech Scientist Podcast.
Want to hear more? Do visit the podcast page and check out other episodes.
Do you wish to simplify your biologics drug development project? Contact Us
