In the rapidly expanding theater of oncology, antibody-drug conjugates (ADCs) stand as the industry’s most ambitious weapon. Often described as "guided missiles" for cancer, these therapeutics combine the precision of monoclonal antibodies with the lethal efficacy of cytotoxic chemotherapy. With approximately 15 FDA-approved products currently on the market and a sector valuation projected to reach $57 billion by 2032, the momentum behind ADCs is undeniable.
However, clinical reality is proving more complex than the initial "targeted chemotherapy" narrative suggested. A significant clinical hurdle remains: many patients whose tumors test positive for a drug’s target fail to respond, or stop responding, to treatment. Gilead Sciences, a major player in the ADC landscape, has partnered with the AI-driven spatial biology firm Nucleai to solve this mystery. By shifting the focus from simple target quantification to a deeper, spatial understanding of the tumor microenvironment (TME), the two companies are attempting to rewrite the rules of precision oncology.
The Disconnect: Why "Target-Positive" Isn’t Always "Target-Accessible"
The fundamental promise of an ADC is to act as a precision delivery system, ferrying a potent payload directly into the heart of a tumor cell while sparing healthy tissue. In theory, if a pathology report confirms the presence of a specific protein—such as HER2—the ADC should successfully bind, internalize, and deliver its toxic cargo.
"Initially, what we were told was that ADCs are basically targeted chemotherapy. I’m going to put a GPS signal on my beachhead… and it’s going to zone in just to the tumor cells," explains Dr. Ken Bloom, head of pathology at Nucleai. "But in practice, it’s far trickier than that."
The "trickiness" stems from a fundamental disconnect between what traditional immunohistochemistry (IHC) assays detect and what the drug actually encounters in the body. Dr. Bloom points to three primary mechanisms of failure:
- Target Inaccessibility: The drug’s target may be present but structurally altered or masked by other molecules.
- Premature Payload Release: Extracellular proteases within the chaotic TME can cleave the ADC’s delicate chemical linkers, causing the payload to release before it reaches the target cell, leading to systemic toxicity rather than tumor death.
- Active Expulsion: Cancer cells are notoriously resilient and can develop drug-efflux pumps that actively expel the ADC payload once it enters the cell.
Chronology of the Partnership
The collaboration between Gilead and Nucleai is the culmination of years of iterative development in digital and computational pathology.
- Early Stages (Pre-2024): The relationship began with limited, focused studies involving multiplex immunofluorescence. These early projects aimed to determine if spatial data could provide more insight than standard clinical snapshots.
- Expansion (2024–2025): Recognizing the potential for data-driven insights, the partnership scaled up. Nucleai began processing whole-slide images (H&E and IHC) from Gilead’s extensive library of clinical trials across various oncology indications.
- August 11, 2026: The companies publicly disclosed the formalization of their expanded partnership, highlighting the use of AI-driven spatial biology to link tissue-level architecture with longitudinal clinical outcomes.
- The Future (Late 2026/Early 2027): Industry analysts expect a peer-reviewed publication of these findings in the coming months, which will likely outline the candidate biomarkers identified through this retrospective analysis.
Supporting Data: The Epitope Problem
One of the most compelling aspects of this research is the "inside-outside epitope problem." For decades, pathologists have relied on IHC staining to identify tumor targets. However, as Dr. Bloom notes, "There’s a reason that we look at the internal side as pathologists… tissue processing tends to preserve the intracellular portions of membrane proteins better."
The problem is that ADCs require an extracellular target to bind. A classic example is HER2. The industry-standard Ventana 4B5 antibody detects the intracellular domain of HER2, which is useful for identifying the presence of the gene. However, many tumors express "p95HER2," a truncated fragment that lacks the extracellular domain that drugs like trastuzumab target.
A 2015 study demonstrated that measuring the intracellular vs. extracellular domain independently led to discordant results in 15% of cases. When these measurements were separated, those with high extracellular-domain expression correlated with significantly longer disease-free survival. Nucleai’s computational tools are designed to look past these historical limitations, analyzing the spatial context of these proteins rather than just their aggregate expression levels.
Official Responses and Strategic Vision
Gilead Sciences views this collaboration as a strategic evolution of its translational medicine and biomarker division. By integrating Nucleai’s computational prowess with its own internal clinical data, Gilead aims to refine its patient selection process.
"Nucleai brings specialized expertise in AI-driven spatial biology and tissue analytics," says Meghna Das Thakur, senior director of oncology biomarkers at Gilead. "Combining those capabilities with our scientific and clinical expertise allows us to generate insights with a level of efficiency that was previously impossible."
For Nucleai, the goal is to transform the pathology report from a static, subjective text document into an objective, data-rich map of the tumor. "We care about the neighborhood that a cell is in," says Dr. Bloom. "What’s the relationship between it and other cells? What’s the density of things? Quantifying those relationships adds a whole new dimension."
Implications for Future Oncology
The implications of this work extend far beyond Gilead’s current pipeline. The industry is currently at a crossroads where the sheer volume of data produced by modern clinical trials has outpaced the human ability to interpret it.
1. Elevating the "Floor" of Pathology
Dr. Bloom argues that AI-assisted pathology acts as an "overreader," catching human errors and standardizing interpretation across different laboratories. By reducing the variability caused by pathologist fatigue or differences in training, AI-driven tools could make clinical trial results more reproducible and reliable.
2. A New Language for Clinicians
One of the most significant barriers in modern oncology is the communication gap between the pathology lab and the bedside oncologist. Computational tools that visualize tumor-immune interactions in color-coded, intuitive spatial maps offer a way to communicate complex molecular data in a format that surgeons and oncologists can immediately act upon.
3. The Path Toward Clinical Adoption
Despite the technological promise, the industry still faces a "missing pathway" problem. There is currently no clear, standardized regulatory route for adopting these complex, AI-driven computational models into routine clinical practice. However, as Dr. Bloom notes, "There are several leaders out there attempting to forge that first path. As soon as the first one hits, you are going to see a wave that follows."
Conclusion: The Next Generation of Precision Medicine
The partnership between Gilead and Nucleai represents a shift from "target-focused" to "context-focused" oncology. While the initial promise of ADCs—that a simple positive biomarker test would guarantee a therapeutic response—has been tempered by the realities of tumor complexity, the integration of spatial biology offers a path forward.
By analyzing the "neighborhood" of the tumor, identifying why certain epitopes remain inaccessible, and using AI to provide consistent, objective data, the industry is moving closer to the true goal of precision medicine: ensuring the right patient receives the right drug at the right time. While the "pathway" for these tools remains under construction, the progress made by Gilead and Nucleai suggests that the next generation of ADC development will be defined not just by the potency of the payload, but by the intelligence of the delivery.
About the Author
