From Consensus Oncology to Individualized Biology

A visualization of liquid biopsy technology, where circulating tumor DNA fragments float in a digital bloodstream. (Credit: Alexander/AdobeStock)

Every year brings new targeted therapies, immunotherapies, and genomic tests, yet far too many patients still cycle through multiple lines of treatment that don’t work, or work only briefly. Behind those statistics is a simple reality: most therapy decisions are still driven by generalized protocols and population-level evidence, not by the unique biology of each patient’s tumor. ¹

At the same time, patients, doctors, and pharmaceutical companies are all searching for the silver bullet: the one drug, process, or technology that will finally solve cancer. While that breakthrough may someday arrive, betting everything on a single answer overlooks what is already within reach. The real opportunity lies at the intersection of biology, technology, artificial intelligence (AI), and drugs working synergistically together, not as isolated components. When live tumor biology, multiomics data, automation, and AI are combined around a single patient, we can start saving lives today while we wait for that proverbial silver bullet.

Why One-Size-Fits-All Protocols Still Drive Cancer Treatment

The dominance of one-size-fits-all regimens is not a failure of intent; it is a by-product of how evidence and guidelines have historically been built. Large clinical trials evaluate therapies in broad, heterogeneous populations, and standard-of-care (SOC) regimens are distilled from what works best on average. This framework has supported safety, consistency, and operational simplicity for decades, and professional groups such as ASCO and the Oncology Nursing Society have invested heavily in standardizing chemotherapy administration to reduce errors. ²

But biology rarely behaves like an average. Systems pharmacology research has shown that variability in drug response and mechanisms of resistance is now one of the primary challenges in oncology. ¹ Even when patients share a diagnosis and biomarker, their tumors can behave very differently. Combination regimens that look superior at the population level often derive their apparent benefit from patient-to-patient variability rather than true synergy in any one individual. ³ In effect, we are often designing therapies to maximize the chance that “someone” will respond, not to guarantee that this person gets the right drug first.

Real-world practice further amplifies this variability. SOC choices can differ across regions and institutions due to access, approvals, and reimbursement. ⁴ Oncologists themselves vary in how aggressively they treat, particularly near the end of life, leading to overuse of low-value chemotherapy for some and underuse of aggressive options for others. ⁵ The net result is that many patients are exposed to toxic regimens that deliver little benefit, while opportunities for better-matched therapies are missed.

Benefits and Limits of Genomics

Genomics has unquestionably advanced cancer care. For some patients, a single driver mutation can unlock a highly effective targeted therapy. But across the broader solid tumor population, the direct benefit of genomics-driven therapy selection remains limited. Analyses of precision oncology programs suggest that only an estimated 2–20 percent of patients with advanced solid tumors derive clear clinical benefit from genomically matched therapies today. ⁶

There are several reasons: not all tumors harbor actionable mutations; resistance pathways evolve quickly; and many “actionable” alterations lack approved drugs or reimbursed access. Even when targeted agents are available, genomic data alone describe what is present in the DNA or RNA, not how the tumor, as a living system, actually responds to a given drug in real time. That missing link is what functional precision medicine aims to provide.

The dominance of one-size-fits-all regimens is not a failure of intent; it is a by-product of how evidence and guidelines have historically been built. (Credit: AdobeStock)

Functional Precision Medicine: Testing Each Patient’s Tumor

Functional precision medicine (FPM) takes a biology-first approach. Instead of inferring likely benefit from other people’s tumors, FPM directly measures how an individual patient’s living tumor cells respond to drugs ex vivo. In a typical workflow, viable tumor cells from a biopsy or surgical specimen are cultured and exposed to a panel of hundreds of FDA-approved oncology drugs and rational combinations at multiple doses. Cell viability readouts create a patient-specific response map that can guide therapy . ^7,8

Cancer Cell and other journals have described FPM as a strategy that can immediately translate into personalized treatment recommendations by identifying vulnerabilities and novel drug combinations. ⁷ Feasibility studies and early clinical trials have reported concordance between ex vivo sensitivity patterns and real-world clinical responses across a diverse patient population and both solid and liquid cancers. ^8-11 FPM has revealed effective off-label options for patients who have exhausted standard therapies. ^10,11

FPM is straightforward: instead of guessing what might work, FPM experimentally measures what actually works on this tumor. Practically, making that FPM scalable and repeatable requires the convergence of several technologies: advanced cell models, robotics, and AI, orchestrated around the individual patient.

AI, LLMs, and Automation: Scaling Individualized Biology

Modern FPM platforms increasingly resemble compact drug-discovery engines centered on one patient at a time. Rapidly cultured patient-derived cells add physiological relevance by better mimicking the architecture and microenvironment of real tumors. ^12,13

These FPM tests generate a torrent of data, leveraging how immune co-culture assays quantify tumor and immune cell behavior under different treatments, while genomic and transcriptomic profiling reveals which pathways may be driving resistance or sensitivity. ^6-8,14 AI models are essential for turning this high-dimensional data into actionable insight. AI/ML algorithms can analyze thousands of data points, classify phenotypic patterns, and distinguish true antitumor activity from nonspecific toxicity. ^12-14 Multiomics models integrate DNA, RNA, and functional readouts to predict which combinations are likely to deliver durable responses versus transient cytotoxicity. ^14,15

Large language models (LLMs) play a complementary role; however, there is a crucial caveat. LLMs and other foundation models are, at heart, consensus engines. They are trained on broad distributions of data and are excellent at predicting what is likely on average. That makes them ideal for analyzing stratified population data and designing systems, but not sufficient for making precise treatment decisions for a single human being. To truly personalize care, AI must be anchored in patient-native data: how this person’s cells respond to specific drugs, what their DNA/RNA shows, and how those signals change over time. Population-trained AI can guide where to look; biology tells us what to do.

The most powerful architectures must pair FPM-generated functional data and genomics with AI and LLM layers: functional and molecular data define the ground truth; AI ranks and explains options; and LLMs contextualize them within guidelines, logistics, and clinical trial landscapes. In that configuration, biology, technology, AI, and drugs stop being separate tools and start operating as a coordinated system around the individual.

Impact on Care Today — and on Tomorrow’s Therapies

In practice, an AI-enabled FPM workflow can be integrated into care in days rather than months. A biopsy or surgical specimen is collected during routine care and shipped to an FPM laboratory. Tumor cells are cultured and distributed robotically across hundreds of wells, then exposed to FDA-approved drugs, emerging agents, and combinations at multiple concentrations. Live-cell imaging, viability assays, and immune co-culture experiments generate phenotypic responses, while genomic profiling characterizes the tumor’s mutational and transcriptional landscape. ^6-8,12-14

Machine-learning models then analyze the dose–response matrix, identify the most effective and least toxic options, and flag nonintuitive combinations. LLMs synthesize these findings with the patient’s treatment history, comorbidities, and guideline-based options to produce a ranked narrative recommendation for the oncologist. ^14-17 The report returns in time to influence the next line of therapy. As treatment is delivered, outcomes and toxicity are captured as real-world evidence (RWE), continuously refining models for future patients. ^3,16,18

Upstream, the same platforms can accelerate therapy development. Because each patient becomes a small, high-throughput experiment, FPM systems can quickly map how classes of drugs perform across genetically and phenotypically diverse tumors. This approach has shown that ex vivo functional data can reveal effective combinations and overcome resistance in hematologic malignancies and solid tumors. ^10,11,19

Thought leaders in drug development have argued that converging pragmatic trials with RWE is essential for advancing precision oncology, especially in rare molecular subsets where conventional trials are slow and underpowered. ¹⁸ AI-enabled FPM offers exactly that kind of data stream.

From Silver Bullets to Synergy: A New Design Mandate

For device engineers and medtech leaders, the takeaway is clear: the biggest breakthrough in cancer care is unlikely to be a single drug or algorithm. It will be the engineered intersection of biology, technology, AI, and therapeutics working synergistically around each individual patient.

Space exploration did not change the laws of physics; companies like SpaceX changed how those laws were used, making launches safer, faster, and cheaper. Oncology is facing a similar moment. We are not redefining biology; we are finally using it to our advantage at the level it actually operates: one person at a time.

That shift demands new design principles. Consensus-based tools (guidelines, pathways, LLMs trained on broad datasets) and individual-anchored tools (FPM assays, patient-native multiomics, AI models trained on those signals) must be treated as complementary but distinct. Consensus is powerful for designing systems, stratifying risk, and planning trials. It becomes dangerous when mistaken for personalization.

If we get this right, the standard question in oncology will change. Instead of asking, “What is the best regimen for patients like this?,” teams will ask, “What does this patient’s own biology tell us about the best regimen, and how quickly can we act on it?” AI-enabled functional precision medicine provides a practical path to that future, improving outcomes today while generating the deep, real-world data needed for tomorrow’s therapies, and bringing us incrementally closer to the long-awaited silver bullet.

This article was written by Jim Foote, Co-founder and CEO, First Ascent Biomedical, Miami, FL. For more information, contact Foote at This email address is being protected from spambots. You need JavaScript enabled to view it. or visit here .