The oncology landscape is currently dominated by the high-stakes evolution of antibody-drug conjugates (ADCs). Often hailed as "guided missiles" of the cancer world, these therapies combine the precision of monoclonal antibodies with the lethal power of cytotoxic chemotherapy. With approximately 15 FDA-approved products and a market valuation projected to reach $57 billion by 2032, the segment is undeniably the most vibrant frontier in drug development.
However, beneath the clinical successes lies a persistent, frustrating mystery: why do some patients whose tumors test "target-positive" for a specific antigen still fail to respond to these sophisticated drugs?
To solve this, Gilead Sciences and AI-powered spatial biology firm Nucleai have embarked on a collaborative effort to bridge the widening gap between traditional pathology and the complex reality of the tumor microenvironment. By analyzing digitized tissue pathology slides, the partners are attempting to determine why some tumors effectively "deflect" treatment, hoping to redefine how we identify patients most likely to benefit from ADC therapy.
The "Target-Positive" Paradox: When GPS Fails
The fundamental promise of an ADC is elegant in its simplicity: a tumor-targeting antibody acts as a GPS, ferrying a potent payload directly to the cancer cell and sparing healthy tissue. "Initially, what we were told was that ADCs are basically targeted chemotherapy," explains Dr. Ken Bloom, head of pathology at Nucleai. "The promise is getting all the benefits of chemotherapy without the toxicity."
In practice, however, the biology is far more treacherous. A tumor may appear to express the target protein on a standard immunohistochemistry (IHC) slide, yet the drug remains ineffective. This resistance stems from several biological factors:
- Epitope Inaccessibility: The specific protein site targeted by the drug may be altered, glycosylated, or physically shielded, rendering it invisible to the ADC.
- The "Inside-Outside" Problem: Most diagnostic antibodies used in pathology laboratories are engineered to bind to the intracellular domain of a receptor, as these regions are easier to preserve during tissue processing. However, ADCs must bind to the extracellular domain to internalize their payload.
- Microenvironment Interference: Extracellular proteases within the tumor microenvironment can prematurely cleave the chemical linkers connecting the drug to the antibody, releasing the toxic payload before it ever reaches its target.
- Cellular Defenses: Cancer cells are notoriously adaptable; many utilize drug-efflux pumps to actively expel the cytotoxic payload once it enters the cell, neutralizing the drug’s effect.
A Chronology of Collaboration
The relationship between Gilead and Nucleai did not begin overnight. It is the result of a multi-year maturation process that has evolved from basic research to deep clinical data mining.
- Early Stages (2022-2023): The collaboration began with a focus on preclinical multiplex immunofluorescence (mIF). The goal was to visualize multiple proteins simultaneously within the tumor microenvironment to understand how cellular neighborhoods influence drug efficacy.
- The Scaling Phase (2024): As both companies recognized the potential of AI to interpret these complex images, the scope expanded. Gilead began providing Nucleai with H&E (hematoxylin and eosin) and IHC whole-slide images from various clinical studies across multiple oncology indications.
- Current Milestone (August 2026): The companies publicly disclosed their expanded partnership. By linking high-resolution tissue features from historical clinical trials with actual patient outcomes, Nucleai is building models that can predict, with greater accuracy than traditional staining, which tumors will respond to Gilead’s pipeline of ADCs.
The Limitations of Conventional Pathology
For decades, the standard for assessing biomarker status has been IHC. While robust for diagnosing cancer, IHC is often insufficient for predicting the efficacy of targeted therapies.
Consider the case of HER2, a classic example in the ADC space. The widely used Ventana 4B5 IHC antibody detects the intracellular domain of the HER2 protein. While this accurately identifies HER2-amplified tumors, it fails to account for "p95HER2"—a truncated form of the protein that lacks the extracellular domain. Because many ADCs, such as trastuzumab deruxtecan, target the extracellular domain, a patient might test "positive" via standard IHC while actually lacking the target required for the drug to bind.
"There’s a reason that we look at the internal side as pathologists, that we try to stay inside the membrane," Dr. Bloom notes. "Tissue processing tends to preserve the intracellular portions better. Yet it’s the outside that you should really care about."
Computational Pathology: The "Neighborhood" Effect
Nucleai’s approach represents a shift from "digital pathology"—which simply digitizes slides for viewing—to "computational pathology," which uses machine learning to extract information invisible to the human eye.
Rather than simply measuring the intensity of a stain (the current standard), the AI platform evaluates the spatial context of the tissue. "Now we care about the neighborhood that a cell is in," says Dr. Bloom. By analyzing the density, spatial distribution, and proximity of tumor cells to other immune or stromal cells, the AI can characterize the "ecosystem" of the tumor.
This spatial information is crucial. A tumor cell that is isolated may interact differently with an ADC than one that is densely packed within a hypoxic, protease-rich environment. By quantifying these cellular neighborhoods, Nucleai and Gilead are identifying candidate biomarkers that are significantly more predictive of clinical outcomes than standard protein expression levels alone.
Official Perspectives: A Strategic Synergy
Gilead Sciences views this collaboration as a strategic evolution of its translational medicine infrastructure. According to Meghna Das Thakur, senior director of oncology biomarkers at Gilead, the goal is to drive precision medicine by integrating deep biological insights into the early stages of clinical trial design.
"Nucleai brings specialized expertise in AI-driven spatial biology and tissue analytics," Das Thakur stated. By combining these insights with Gilead’s deep clinical datasets, the company aims to move beyond trial-and-error medicine. The objective is to identify the "signatures of resistance" early, allowing Gilead to refine its patient selection criteria and potentially salvage ADCs that might otherwise be dismissed as ineffective.
The Future of the "AI Overreader"
While the technology is transformative, Dr. Bloom is quick to caution against the total removal of human judgment. He emphasizes the concept of the "AI overreader"—a tool that assists the pathologist rather than replacing them.
"A fool with a tool is still a fool," Bloom says, referencing a long-standing mantra from his teaching career. He argues that AI is best utilized to:
- Standardize Interpretation: Reduce the inter-observer variability that often plagues pathology, where two experts might read the same slide differently due to fatigue or training differences.
- Improve Communication: Computational tools can generate intuitive, color-coded maps of tumor microenvironments that are far easier for oncologists and surgeons to interpret than dense, text-heavy pathology reports.
- Enhance Accuracy: By catching subtle nuances in tissue architecture, the AI acts as a safety net, ensuring that no clinically relevant signal is overlooked.
Broader Implications for the Industry
The success of this collaboration could set a new industry standard. Currently, there is no "tried-and-true" regulatory or technical pathway for using computational pathology biomarkers in drug labeling. Most companies are operating in a "wild west" of innovation, attempting to forge their own paths through the FDA and other regulatory bodies.
"I think there still isn’t a pathway that’s been forged clearly yet for somebody else to follow," Dr. Bloom acknowledges. "But the good news is that there are several leaders out there attempting to forge that first path. What you’re going to see, as soon as the first one hits, is a wave that follows."
As Gilead and Nucleai prepare to release their findings in upcoming peer-reviewed publications, the industry is watching closely. If they can successfully prove that spatial AI can identify responders where standard IHC fails, it will likely trigger a massive shift in how clinical trials are designed, how drugs are approved, and ultimately, how oncologists choose the right therapy for the right patient.
For the $57 billion ADC market, the solution to the "resistance problem" may not lie in better chemistry, but in better biology—seen through the lens of artificial intelligence.
About the Author
