Picture a familiar scenario: a promising cell therapy program advances through early clinical development on the strength of a phenotypic marker panel and a release assay that was never meant to do more than confirm identity. Then comes the process change, the site transfer, or the pivotal trial, and the analytics that quietly carried the program are suddenly asked to carry a regulatory submission. The assay doesn’t transfer cleanly. The comparability package is thin because there was never a meaningful historical dataset behind it. And the potency claim, built on marker expression rather than a demonstrated mechanism of action, draws exactly the kind of question no one wants at that stage.
This story plays out across the field more often than it should, and it isn’t really a story about assay failure. It’s a story about timing. Programs that build advanced analytics into their earliest development decisions, rather than treating them as a late-stage deliverable, end up in a fundamentally different position when it matters most.
Potency Is a Decision, Not a Deliverable
The first reframe worth making is a simple one: identity and potency are not the same thing, and marker expression alone can only ever tell you the former. A surface marker, a transcript, or a phenotypic signature tells you what a cell population is. It doesn’t tell you what that population will do once it’s infused into a patient. Potency, by definition, has to be tied to function, and function has to be tied to mechanism of action.
Regulators have been moving in this direction as well. FDA’s draft guidance on potency assurance for cellular and gene therapy products, issued in late 2023, explicitly encourages sponsors to move away from single surrogate measures and toward a matrix of attributes anchored to the product’s mechanism of action, particularly as clinical and manufacturing experience accumulates over a product’s lifecycle. That’s a meaningful signal to programs still leaning on a single marker-based potency assay: the expectation is shifting toward functional, MOA-linked characterization, and toward getting there sooner rather than later.
Three Payoffs of Building Analytics In Early
When advanced analytics are treated as a foundational part of process development rather than a late add-on, three benefits tend to compound over the life of a program.
Tech transfer becomes routine instead of a fire drill. Assays that are qualified and stress-tested early, with reference standards established from the outset and methods designed to be as platform-agnostic as practical, move cleanly between process development, CDMOs, and GMP manufacturing sites. Assays retrofitted under deadline pressure, by contrast, tend to carry undocumented assumptions that only surface once they’re in someone else’s hands.
Comparability gets a real foundation instead of a thin one. Every program eventually changes something: scale, site, raw material, or process step. Comparability protocols are only as strong as the historical dataset behind them, and that dataset can only be as long as the analytics have been running. Starting advanced characterization early, even in preclinical or IND-enabling work, means that by the time a change needs to be justified, there’s a real trend to justify it with.
Potency claims hold up under scrutiny. A function-based, MOA-anchored assay, built and refined with a program’s biology from the start, tends to remain scientifically defensible as understanding of the product matures. A marker-based surrogate adopted for convenience early on often becomes the hardest thing to defend later, precisely because it was never meant to answer the question being asked of it at BLA.
Advanced Analytics as the Enabler
None of this is an argument for complexity for its own sake. It is an argument for using tools that make function-based, multi-attribute characterization practical rather than aspirational. Technologies such as real-time label-free laser force cytology, multiparametric flow cytometry, high-content imaging, single-cell multi-omics, cytokine and secretome profiling, and increasingly machine learning-driven data integration make it possible to develop a more complete and functional understanding of a cell product. Together, these approaches allow developers to move beyond static measurements and build analytical strategies that better reflect how cells are likely to perform in patients. In doing so, they transform the principle that potency should reflect function into something that can be measured, monitored, and applied consistently across batches, sites, and manufacturing campaigns.
Where to Start
The practical version of this advice is straightforward, even if the execution takes discipline. Bring the analytical and bioanalytical function into the room at the preclinical or IND-enabling stage, not after Phase 1 data are in hand. Treat the potency assay as a living, MOA-linked tool that gets refined as understanding of the product deepens, rather than a static checkbox set once and revisited only under pressure. And build the data package with an eye toward where it will need to travel, so that tech transfer, comparability, and potency are being designed for from day one rather than reconstructed after the fact.
Programs that make this investment early don’t just avoid a painful retrofit later. They put themselves in a position where the science, not the calendar, is driving the decisions that matter most to patients.
By Renee Hart, President and Co-founder, LumaCyte
Sources
Potency Assurance for Cellular and Gene Therapy Products — FDA guidance document
