Antibody-drug conjugates have transformed oncology over the past decade, but they have been working with a limited playbook. The same small set of validated ADCs — including HER2, TROP2, NECTIN4, cMET and a few others — has anchored most ADC programs, not because researchers lack ambition, but because the biology of what makes a safe ADC target is genuinely constrained.
Promatix Biosciences, a UK-based biotech, is attempting to expand that constraint using a proteomics-driven cis-bispecific discovery platform. By identifying pairs of antigens that are co-expressed on cancer cells but rarely co-expressed in normal tissue, the company aims to move ADC development from a field limited by target scarcity to one enriched with first-in-class target combinations to evaluate.
Why target selection became the bottleneck
For most of ADC development's history, target selection was not the primary problem. "The need for targets that are differentially expressed on cancer cells has until now been eclipsed by the need to reduce the non-specific toxicity associated with ADCs that severely limits their therapeutic window," Mark Edwards, Chief Technology Officer at Promatix, told DDN.
Improvements in antibody selection, conjugation technology, and linker-payload structures over the past two decades have substantially addressed those toxicity problems. That progress, however, has raised the bar on what counts as an acceptable target. A drug with a well-controlled payload now needs a highly selective delivery mechanism, and very few individual membrane proteins are expressed exclusively on cancer cells. The result is a narrowed target landscape precisely at the moment when better payloads could support more aggressive programs.
The logic-gate approach
Promatix's solution is to shift the unit of targeting from a single antigen to a pair. Whole proteome analysis of the cancer surfaceome reveals that truly cancer-selective single markers are rare. But pairs of proteins that are co-expressed on cancer cells while having non-overlapping expression in normal tissue are far more common — and the platform is designed to identify them at scale.
The mechanism that allows this co-expression pattern to be exploited is avidity-driven dual engagement. Antibodies are bivalent by design, with two binding arms, and the binding affinity they achieve when engaging a polyvalent target — a cell surface displaying both antigens — is 100 to 1,000 times higher than monovalent binding. By engineering a bispecific antibody where each arm recognizes a different surface protein, the platform creates a targeting agent that achieves high-affinity binding only on cancer cells that express both antigens simultaneously, with dramatically reduced binding to normal tissue where only one antigen is present.
Early validation in colorectal cancer
The company's lead program targets EGFR and EphA2 in colorectal cancer — a combination chosen because the two antigens are co-expressed in tumor tissue but have distinct expression profiles in the normal tissues most at risk from EGFR- and EphA2-directed toxicity: keratinocytes and endothelial cells respectively.
Preclinical evaluation showed that the bispecific retained potent activity against colon cancer cell lines in vitro and in vivo while sparing those normal cell populations. The program also demonstrated a higher degree of internalization compared to bivalent monospecific antibodies targeting either antigen alone — a finding that suggests improved payload delivery to tumor cells and a correspondingly better therapeutic index. The lead candidate is now in animal studies assessing tolerability.
From target scarcity to an embarrassment of riches
The platform's broader implication is a shift in how the ADC field approaches pipeline construction. Where target identification has historically been a limiting step, the bispecific approach opens up what Edwards described as "an embarrassment of riches" — a large number of predicted tumor-selective antigen pairs that now require triage.
The process for prioritizing combinations runs through three stages: in silico cross-checking of target expression using proteomics and RNA data alongside factors such as protein topology, shedding, and internalisation rates; confirmation of co-expression and co-localization in cancer biopsies and cell lines using immunohistochemistry and immunofluorescence; and, for leads that emerge from that process, functional evaluation of bispecific panels carrying binders with different affinities.
The question now being tested is whether the expanded target landscape translates into a correspondingly expanded set of viable ADC programs — and whether the logic-gating principle holds as the platform moves from colorectal cancer into other solid tumor indications.
