The high cost and high failure rate of drug development is a well-documented crisis in biomedicine. A significant contributing factor is the heavy reliance on animal models that often fail to accurately predict human responses, leading to late-stage clinical trial failures despite promising preclinical results.
In response, a profound scientific and regulatory shift is underway, moving the industry toward a human-centered approach. Spearheaded by initiatives like the US Food and Drug Administration’s (FDA) 2025 framework for New Approach Methodologies (NAMs), this change redefines “acceptable evidence,” prioritizing human-relevant data from in vitro systems, advanced organ models, and in silico simulations from the very beginning of the discovery process.
This transition is not merely technological, as it represents a fundamental change in philosophy. It promises to make drug development faster, cheaper, and more predictive, while also addressing long-standing issues of inclusivity and equity in clinical research by anchoring science in diverse human biology from the outset.
To explore the implications of this shift, the Drug and Device World spoke with Dr Greg Tietjen, co-founder and CEO of Revalia Bio. With a unique background that spans biophysics, nanomedicine, and a former faculty position at Yale, Dr. Tietjen is at the forefront of building the infrastructure needed to support this human-centered future. In the following interview, he discusses the regulatory momentum, the integrated future of research models, and the potential for a more ethical and effective path to new therapies.
This interview has been edited for clarity, consistency, and length.
Phalguni Deswal (PD): The FDA’s 2025 framework recognizes a shift away from traditional animal models. What is the most important result of this redefinition of “acceptable evidence”?
Greg Tietjen (GT): The biggest thing is a full community commitment to human-centered development. This doesn’t mean animal models don’t have a role, but it means we start from an anchoring in understanding human biology—we don’t get there at the end, which was the old paradigm. It’s created a “permission structure” for the industry. For a long time, it felt like we were checking boxes that weren’t always relevant because we had to do something. Now, with commitments to reducing animal models and the NIH requiring human data in earlier research, we can explore new ways to start with the human center, which is our ultimate goal.
This creates urgency and opportunity, but also some anxiety. The challenge is that human-centered science depends on access to human tissues and data, which isn’t as simple as ordering from a catalog. It requires a whole new expertise and infrastructure. We believe the future is a new model, similar to the cloud computing revolution, where developers have on-demand access to regulatory-grade human data infrastructure, allowing them to focus on developing therapeutics, not building and maintaining the tools for development.
PD: With this shift to human models, what new questions can we start asking at the very beginning of drug development?
GT: Interestingly, the foundational questions remain the same: How does the disease work? What drug do I choose? Can I get the drug to the right site? Is it toxic elsewhere? Is it effective once it gets there? What’s changing is that we are now improving our ability to have more complete answers.
I heard a beautiful metaphor from an FDA leader: reviewing current investigational new drug (IND) packages can be like reading a novel with chapters ripped out. You might have good data from animals where the biology is preserved, but you’re missing key parts of the story. For example, in cancer, we’ve long used mouse models where tumors grow in the flank. These are poor models for the actual delivery environment of a human tumor, so that chapter on “can I deliver my drug effectively?” is missing.
With an integrated approach, we can fill in those chapters. At Revalia, through partnerships with organ procurement organizations and academic healthcare systems, we can, for instance, bring a non-transplantable, tumor-bearing human lung back to life on a perfusion system. We can make it breathe, use clinical-grade imaging, and perform an image-guided drug delivery study to answer that specific question. By layering different models, we can tell a more complete story, making clinical trials far less risky and more likely to succeed.
PD: A lack of diversity and inclusivity has long plagued clinical trials. Does this human-first approach allow for better representation of diverse populations earlier in the process?
GT: Certainly. By starting science at the source of the unmet need, with donated tissues from patients we haven’t been able to save, diversity is baked in. Instead of creating contrived models to try and recreate a diverse patient population, we are working with the actual patient populations that are hardest hit by the disease. We have a partnership in Oklahoma with LifeShare of Oklahoma, for instance, where there’s a high incidence of lung cancer. We’re working with the patient population that we can’t treat today, and their gifts through organ donation for research create the opportunity to understand their disease in a direct, not contrived, way.
This approach ensures we are anchored in the diverse patient populations we’re trying to develop treatments for. It also means we don’t get fooled by small effect sizes in constrained populations; we start with diversity and look for big signals. This honors the profound gift of donation, providing agency and legacy for donors and their families while ensuring our research is grounded in the reality of human disease across all communities.
PD: Artificial intelligence is a major topic in pharma. What early “wins” could help build trust in AI and human-data-driven approaches?
GT: First, we must level-set: we don’t yet have the integrated “internet of data” for human health needed to train all-powerful AI models for drug development. We need a more open-source, collaborative mindset to build that resource because no single organization will be capable of generating the necessary data.
The first use cases will likely build on where LLMs are already strong: helping scientists collaborate and manage complexity. But for the future of AI in drug development, I look to the toxicity space. This is a well-defined question: “If my drug goes where I don’t want it, is it toxic?” The opportunity here is to build “digital twins”, which are digital replicas of human systems and toxicity pathways powered by machine learning. We could then layer predictive AI tools on top to screen for these risks.
The key is stepwise progress. We’ve shown AI/ML models can discover drugs. The next win is showing that it can predict toxicity. Then, predict effective delivery. Then, predict efficacy. By finding these wins step-by-step on specific drugs in collaboration with the FDA, we can run the “first four-minute mile,” and then the floodgates will open.
PD: What are the key takeaways from this ongoing shift in drug development?
GT: The future is human-centered, and we stand on the shoulders of all the work that came before. But the biggest conceptual takeaway is that we must get to a place where failing a human experiment is no longer a catastrophic event, as it is in a failed clinical trial, but rather a catalytic engine for learning. That’s the heart of innovation.
Think about software technology versus CRISPR. Software dominates our lives because it can fail fast without dire consequences, while CRISPR, born around the same time as the iPhone, is only now yielding clinical results. We have to adopt a new mindset where failure in human model systems becomes a catalyst for innovation, not a tragedy to be avoided at all costs. That is the true paradigm shift.
