Cell and gene therapy (CGT) has arrived. The question a panel at the Convention on Pharmaceutical Ingredients (CPHI) Americas 2026 in Philadelphia spent an hour working through was not whether these modalities work — it was whether the industry surrounding them is built for what comes next.
Moderated by Bernardo Estupiñon, the session brought together Bruce Thompson, President and Chief Technology Officer of Kincell Bio; Gregory MacMichael, President of CMC Bioservices and former Global Head of Cell and Gene Therapies at Novartis; Irene Rombel, Chief Executive Officer of BioCurie; and Kenneth Yancey, Senior Vice President, Head of CMC at Franklin Biolabs. Together they mapped a field at a genuine inflection point, where the biology has outpaced the infrastructure built to deliver it, and where the path forward requires rethinking how discovery, development, and manufacturing relate to each other from the very beginning.
The third wave of medicine and what it demands
Thompson framed the current moment in historical terms. “Small molecules were the first wave of modern medicine. Large molecule biologics were the second. CGTs represent a third wave,” and the lesson from the first two, he argued, is that each wave requires a fundamentally different approach to how therapies are designed and built.
The implication for CGT is that manufacturing cannot be an afterthought. Thompson made the case that development teams need to work backward from the target product profile — the indication, the dose, the patient population — all the way through to the manufacturing process, before committing to a particular biological approach. The discipline of building the delivery system before the therapy is fully defined is, in his view, what distinguishes a program that reaches patients from one that stalls in development. The field has a tendency to overshoot its indication early on, he noted, and as the biology becomes better understood, programs can be pulled back to a more precise target. But that recalibration is much harder, and expensive, if the manufacturing assumptions have already been set.
He also raised a pointed observation about AI's role in this process: Predictive tools work by pattern recognition, and they are only as reliable as the data they have been trained on. For a field as young as CGT, where the accumulated data from clinical programs is still thin compared to small molecules or conventional biologics, AI tools carry inherent limitations that practitioners need to understand before relying on their outputs.
Durability, cost, and the clinical feedback gap
MacMichael brought the conversation to one of the field's most persistent and commercially consequential problems, the gap between what CGT can achieve in principle and how long it actually lasts in patients.
Many cell therapy recipients, he noted, see meaningful benefit for six months to a year — and then the effect wanes. A significant driver of that durability problem is antigenic escape, the mechanism by which tumors evolve to evade the engineered immune cells that were designed to eliminate them. That is a biological problem, and solving it requires learning from clinical experience — from the programs that failed as much as from those that succeeded.
The challenge, MacMichael observed, is that the clinical feedback loop in CGT is broken in a way it is not in other therapeutic areas. When a drug fails in a clinical trial, the field can examine what went wrong and apply those lessons to the next compound. In CGT, regulatory and intellectual property structures mean that learnings from one company's program are rarely transferable to others, even when the underlying biology is shared. The field is generating clinical data that could inform the next generation of products, but the mechanisms for sharing that knowledge across the industry in a way that accelerates development for everyone do not yet exist in any systematic form.
On AI specifically, MacMichael's view was direct: The window for shaping how these tools are adopted in CGT development is closing. The field needs to engage with AI actively and critically now, rather than waiting for the tools to mature further, or risk having the direction of their development set by others.
Building for what the tools can't yet do
Rombel offered the session's most technically granular perspective, and the one that pushed hardest against the assumption that existing frameworks are adequate for the problems CGT presents.
The field, she argued, made a foundational error early on by treating CGT as a biologics problem — importing the tools, models, and mental frameworks developed for antibodies and therapeutic proteins into a domain where the complexity is categorically different. Vectors, gene editing systems, and cell-based products operate through mechanisms that those frameworks were not designed to characterize. The tools need to be reinvented, not adapted — and that requires genuine innovation, not incremental refinement of what the biologic space already has.
Rombel was particularly focused on AI. The technology is not a solution in itself; it is a set of capabilities that require domain expertise to apply correctly. Sophisticated, nuanced models built on deep scientific understanding of the underlying biology are what the field needs — not general-purpose machine learning layered on top of inadequate data. Highly educated people, she noted, can be misled by AI-generated outputs that appear authoritative but reflect gaps in training data or model architecture. Critical thinking about what AI tools are actually doing, and what their limitations are, is not a skill the field can afford to skip.
Her framing of success was instructive: not one modality winning out over others, but “multiple modalities, viral vector gene therapies, nucleic acids, cell therapies, and combinations, working together based on comparative advantage, matched to the indication they are best suited to address.” The question of which approach is right for a rare disease affecting a handful of patients and which is right for a condition like Alzheimer's disease are fundamentally different questions, and the field needs the maturity to treat them that way.
On process optimization, Rombel stressed that “the time to get a manufacturing process close to optimal is at the beginning, not after clinical development has generated data with a suboptimal process.” Because there is limited room to make major process modifications once a product is in clinical trials, Rombel noted, and even less once it is approaching approval. The discipline of front-loading process development, even when the biology is still being worked out, is what creates the commercial headroom to actually deliver a product at scale.
Knowledge, patients, and the path to mainstream medicine
Yancey closed the scientific portion of the discussion with a perspective rooted in patient biology. The deeper the understanding of the individual patient, their disease mechanism, their immune environment, their specific genetic variant, the more precisely a therapy can be designed to address it. The field has evolved past the application of broad cures, he argued, toward a more targeted model in which the specificity of the intervention is matched to the specificity of the disease.
He also identified a structural gap in the industry's collective knowledge base: Companies working with CGT platforms tend to accumulate experience that is product-specific rather than platform-wide, and that experience does not systematically transfer across programs or organizations. Consultants and cross-product collaborations serve a function here that internal teams alone cannot — bringing the breadth of cross-program learning that no single company, working on its own pipeline, can accumulate. Sharing what is platform behavior versus what is product-specific behavior is, in his view, one of the most underleveraged opportunities in the field.
The panel's closing consensus pointed toward a horizon that was optimistic but clear-eyed. The holy grail — moving from autologous therapies, which are manufactured individually for each patient, to allogeneic off-the-shelf products — remains a defining goal. Getting there requires not just scientific advances in immune tolerance and cell engineering but the manufacturing systems, quality frameworks, and regulatory precedents that make scalable production possible. The industry has a lot of learning still to do, and much of it will happen in the clinic, in the manufacturing suite, and in the accumulated experience of programs that have not yet been run.
