The landscape of modern oncology is currently dominated by a singular, high-stakes ambition: the refinement of antibody-drug conjugates (ADCs). Often described as “guided missiles” of the pharmaceutical world, these sophisticated molecules are designed to deliver lethal cytotoxic payloads directly to tumor cells while sparing healthy tissue. With 15 FDA-approved products currently on the market and a sector valuation projected to soar to $57 billion by 2032, ADCs represent the frontier of precision medicine.
However, clinical reality has exposed a troubling discrepancy. Many patients whose tumors test positive for the target protein—the "beachhead" for the drug—still fail to respond to treatment. To address this, Gilead Sciences has entered a strategic partnership with Nucleai, a leader in AI-powered spatial biology, to investigate why these target-positive tumors exhibit resistance. By applying advanced computational pathology to clinical trial datasets, the two companies are attempting to decode the hidden mechanics of the tumor microenvironment that render standard diagnostic tools insufficient.
The Core Problem: Why “Target-Positive” Doesn’t Always Mean "Drug-Sensitive"
The fundamental promise of an ADC is simple: an antibody latches onto a specific cell-surface protein, the cell internalizes the complex, and the cytotoxic payload is released, triggering cell death. Yet, as Dr. Ken Bloom, head of pathology at Nucleai, observes, the process is far more complex in the human body than in a laboratory beaker.
“Initially, what we were told was that ADCs are basically targeted chemotherapy,” Bloom explains. “The promise is getting all the benefits of chemotherapy without the toxicity. But in practice, it’s far trickier than that.”
Resistance occurs through several sophisticated mechanisms. First, the cell-surface target may be physically altered or masked, rendering it inaccessible to the antibody. Second, the tumor microenvironment is often a hostile, chaotic space filled with extracellular proteases capable of cleaving ADC linkers prematurely, causing the payload to leak before it ever reaches its intended target. Finally, cancer cells are adept at survival, often upregulating drug-efflux pumps that actively expel the cytotoxic payload once it enters the cell.
The "Inside-Outside" Epitope Paradox
A critical hurdle in current diagnostic practice is the nature of immunohistochemistry (IHC) assays. Most standard pathology antibodies are designed to bind to the intracellular (inside) domain of a protein because those regions are better preserved during standard tissue processing.
However, the ADC itself must bind to the extracellular (outside) domain to initiate the internalization process. In cases like HER2-positive breast cancer, this creates a "hidden secret" in pathology: we are measuring the presence of a protein by looking at a domain the drug doesn’t actually see. If a tumor expresses a truncated version of the protein that lacks the extracellular docking site, an IHC test will report the tumor as "positive," even though the ADC has no functional way to engage the cancer cell.
Chronology of a Collaboration: From Preclinical to Clinical Insight
The partnership between Gilead and Nucleai is the culmination of years of iterative research. The relationship began in the preclinical space, with Nucleai providing multiplex immunofluorescence (mIF) analysis to help Gilead’s scientists understand the spatial architecture of tumor samples.
- Initial Phase (Pre-2024): The companies focused on exploratory work, utilizing small-scale tissue studies to establish baseline correlations between protein expression and drug behavior.
- Expansion (2024-2025): Recognizing the limitations of manual pathology, the collaboration shifted toward large-scale retrospective analysis. Gilead granted Nucleai access to H&E (hematoxylin and eosin) and IHC whole-slide images from multiple clinical trials across various oncology indications.
- The Disclosure (August 2026): The companies officially announced the expansion of their partnership, confirming that they are now using AI to link granular tissue features—such as cell density and spatial distribution—with actual clinical outcomes observed in patients.
Supporting Data: Moving Beyond Standard Expression Levels
The collaboration’s preliminary findings suggest that standard quantification—simply counting how much of a protein is present—is an outdated metric for predicting ADC efficacy.
“What fell out of the analysis was several candidate biomarkers that are more predictive than standard immunohistochemistry,” Bloom says. These candidates are not new proteins or genetic mutations, but rather spatial patterns—the "neighborhoods" in which cells reside.
By using computational pathology, Nucleai’s AI can map the tumor microenvironment at a level of resolution impossible for the human eye. This includes:
- Spatial Distribution: How close are the target-positive cells to the vascular supply or the immune cells of the host?
- Cellular Neighborhoods: Does the presence of specific stromal cells around a tumor cell inhibit or facilitate drug delivery?
- Protein Accessibility: Using spatial markers to determine if the extracellular domain is actually exposed on the cell surface, rather than just identifying the presence of the intracellular domain.
These factors provide a multidimensional view of the tumor that explains why two patients with identical "HER2-positive" scores can have vastly different clinical responses to the same ADC treatment.
Official Perspectives: The Role of the Physician in the Age of AI
Gilead Sciences views this collaboration as a natural evolution of its translational medicine capabilities. According to Meghna Das Thakur, senior director of oncology biomarkers at Gilead, the goal is to integrate AI-driven spatial biology into the drug development pipeline to gain insights more efficiently. By combining Gilead’s deep expertise in drug-target biology with Nucleai’s high-throughput tissue analytics, the company aims to "de-risk" its clinical trials and better identify the patient populations most likely to benefit from its pipeline.
The "Overreader" Model
Dr. Bloom emphasizes that the goal is not to replace the pathologist, but to augment them. He draws a firm line between digital pathology (which digitizes images) and computational pathology (which interprets them).
“A fool with a tool is still a fool,” Bloom remarks, underscoring his belief that AI must be guided by expert medical judgment. He posits that computational tools act as an "AI overreader," reducing the fatigue and subjectivity that naturally occur when a human pathologist reviews hundreds of slides. In his view, the most powerful diagnostic entity is a "well-trained physician with AI," a combination that is "light years ahead" of either a lone human or a standalone algorithm.
Implications for the Future of Oncology
The implications of this work extend far beyond the Gilead-Nucleai partnership. If successful, the development of these "spatial biomarkers" could revolutionize clinical trial design. Instead of enrolling patients based on a binary "positive/negative" IHC score, pharmaceutical companies could screen for "spatially optimized" patients, significantly increasing the probability of trial success and, ultimately, patient survival.
Furthermore, this shift could change the very nature of the pathology report. Currently, reports are often dense with narrative text that can be difficult for oncologists to translate into treatment decisions. Computational pathology promises a more intuitive, visual output: color-coded maps of the tumor microenvironment that provide oncologists with a clear, actionable picture of the tumor’s landscape.
A New Path for the Industry
While the industry is still in the "forging" phase—lacking a singular, standardized pathway for implementing these AI-driven diagnostic tools—the momentum is undeniable. Gilead and Nucleai are among the leaders attempting to build this infrastructure.
As Bloom predicts, the industry is waiting for a breakthrough. "I think there still isn’t a pathway that’s been forged clearly yet for somebody else to follow," he notes. "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."
In the race to master ADCs, the winning strategy may not be to design a more powerful drug, but to better understand the complex, shifting terrain of the tumor itself. By turning the microscope into a computational powerhouse, Gilead and Nucleai are moving toward a future where "target-positive" is no longer a guess, but a precise, evidence-based prediction of success.
About the Author
