Dan Klevisha, CEO, Pion looks at why predictive science moves us beyond educated guesswork.
Imagine investing five years optimising a formulation only to discover in first-in-human studies that its bioavailability
Pion
makes commercial success impossible. For many formulation teams, that possibility is not hypothetical, it's an accepted reality of drug development.
For decades, formulation scientists have accomplished remarkable things despite working with one significant disadvantage: they often make critical development decisions without knowing how a formulation will ultimately behave in humans.
This has never been a reflection of scientific capability or rigor. Rather, it reflects the limitations of the tools available. Throughout much of pharmaceutical development, formulation teams rely on a combination of in vitro testing, animal models, historical experience and informed judgment to guide candidate selection. Yet the definitive answer, human bioavailability, typically remains unknown until years later, when clinical studies finally begin. This creates one of the industry's largest and most accepted blind spots.
As drug molecules become more complex and delivery strategies continue to evolve, particularly with long-acting injectables, biologics and the rapid expansion of subcutaneous therapies, that uncertainty becomes increasingly expensive. Every formulation decision influences development timelines, manufacturing strategy, dosing, commercial viability and ultimately whether a promising therapy reaches patients.
The question facing the industry is now how much earlier we can begin reducing uncertainty.
The cost of waiting
Modern pharmaceutical development is built on thoughtful, evidence-based decision making. Yet formulation teams routinely move programs forward with significant unknowns.
A development group may spend years optimising a formulation only to discover during clinical testing that bioavailability is insufficient to achieve therapeutic efficacy. At that point, options become limited. Concentration requirements, injection volumes, manufacturing constraints and cost-of-goods may make reformulation impractical. In other words, some of the most consequential formulation questions cannot be answered until substantial investments have already been made.
Historically, animal studies have helped bridge this gap, but everyone in the industry recognises their limitations. Biological variability, species differences and poor translation to human performance often make these studies directional rather than predictive. They remain an important part of development today, but few scientists would argue they consistently provide the answers formulation teams ultimately need.
The result is today’s reality: many formulation decisions are still made using the best information available rather than the information developers would love to have in hand.
From measuring to predicting
A transformation is now taking place across pharmaceutical R&D. The industry's focus is shifting from measuring experimental outcomes to predicting them.
Driven by advances in AI, this shift also benefits from mechanistic models, and physiologically relevant in vitro systems, advanced computational approaches and increasingly sophisticated datasets. These tools work together to answer more probing scientific questions and provide earlier insight into drug performance.
These technologies should not be viewed as competing approaches. Instead, they represent complementary tools that help answer different scientific questions. Mechanistic models explain why a system behaves as it does by incorporating known biological and physicochemical processes. AI identifies patterns that may not be immediately obvious within large datasets. Together, these approaches can dramatically improve confidence in formulation decisions long before first-in-human studies occur.
The common denominator across these approaches is the need for data. Lots of it. High-quality experimental data remain the foundation upon which every predictive model depends. AI cannot compensate for poor inputs, and mechanistic models cannot correct for missing information. Better predictions require better measurements.
This is where advanced formulation analytics become increasingly valuable, not simply because they characterise formulations, but because they generate the physiologically relevant data needed to make predictive tools more accurate.
The rise of new approach methodologies
The movement toward predictive formulation science is occurring alongside another important industry trend: the growing adoption of New Approach Methodologies (NAMs).
Regulators around the world are actively encouraging alternatives that reduce dependence on animal testing while improving scientific relevance. Academic institutions are investing heavily in predictive technologies. AI-enabled drug development continues to accelerate. Organ-on-chip platforms, digital models and advanced in vitro systems are steadily becoming part of mainstream research conversations rather than future concepts.
We are headed toward an integrated ecosystem that allows researchers to generate better evidence earlier in development. For formulation scientists, this creates an opportunity to rethink longstanding workflows. Instead of waiting years to understand whether a formulation is likely to succeed, predictive tools can help identify promising candidates earlier, eliminate weaker options sooner and focus resources where they have the greatest likelihood of success. And reducing uncertainty to any extent can have an outsized impact across an entire development program.
Better decisions, not perfect predictions
Predictive science will never eliminate every unknown because biology is simply too complex. And prediction should not replace experimentation; laboratory studies will always remain essential.
Better prediction capabilities will allow organisations to ask better questions, prioritise better candidates and make better decisions throughout development. Sometimes the answer will confirm assumptions and experience, and other times, it will challenge them. Both outcomes provide value if it occurs early enough to influence development strategy.
Perhaps most importantly, predictive approaches acknowledge something formulation scientists have understood all along – that every additional piece of meaningful data helps narrow uncertainty and improve decision quality.
Building the future of formulation science
The pharmaceutical industry is at an exciting inflection point. Computational modelling, AI, advanced analytical technologies and physiologically relevant experimental systems are converging in ways that were not possible even a few years ago. Individually, each technology provides incremental insight. Together, they offer the possibility of fundamentally changing how formulation decisions are made. Organisations that embrace this shift can reduce development risk and the reliance on educated guesswork.
Ultimately, the future of formulation science is not about replacing scientists with algorithms or laboratory experiments with computer models. It is about equipping formulation teams with better evidence earlier, enabling them to move beyond educated guesswork toward informed prediction.
The shift from accepting uncertainty to actively reducing it may prove to be one of the most important advances in pharmaceutical formulation over the next decade.