Despite the improved specificity and efficacy of novel antibody-drug conjugates (ADCs) targeting HER2, such as T-DXd and T-DM1, it is still unclear which patients actually benefit from these therapeutics (1-3). Without better biomarkers for improved stratification, patients may suffer side effects, such as lung toxicity, without therapeutic benefits.
A new study published in Cancer Cell shows that combining spatial biomarkers with AI-based predictive models can reveal new biomarkers of efficacy, helping identify patients most likely to respond to next-generation HER2-targeted ADCs.
Starting Point: A Novel anti-HER2 ADC
The authors performed a translational study based on the HER2+ breast cancer cohort of the phase 2 FASCINATE-N trial (NCT05582499), which included 60 patients in all. The monotherapy arm employed the novel ADC, SHR-A1811 (trastuzumab conjugated to a novel topoisomerase I inhibitor payload, SHR9265, via a degradable linker).
Biopsy samples were analyzed using DNA and RNA sequencing, computational pathology (on H&E-stained and HER2 IHC-stained whole-slide images), and single-cell in situ spatial imaging (10x Xenium). Efficacy was measured as pathological complete response (pCR).
Hormone Receptor (HR) Status Dictates Distinct Response Mechanisms and Biomarkers
While the overall pCR rate of the entire cohort was 63.3%, dividing the cohort revealed a critical need to stratify patients by HR expression status. HR− patients showed a significantly higher pCR rate (75.8%) compared to HR+ patients (48.1%).
- HR- Tumors: Efficacy is linked to tumor-immune microenvironment. Spatial biomarkers of efficacy in HR- tumors clustered around the distribution of immune cells, such as tumor-infiltrating lymphocytes (TILs). Computational pathology, single-cell spatial imaging, and quantitative immunohistochemistry revealed that the degree of infiltration by cytotoxic T cells within tumors was most strongly associated with response.
- HR+ Tumors: Efficacy is linked to spatial organization of HER2+ tumor cells. Surprisingly, pCR rate in HR+ tumors was correlated with how closely HER2+ tumor cells were clustered together. A uniform distribution of HER2-strong-positive cells was associated with higher pCR rate, while tumors with aggregated clusters of HER2-strong-positive cells showed lower response.
Gene Expression Signatures of Response in HR2+ Tumors
The authors mined the single-cell gene expression data further to determine molecular determinants associated with tumor cell aggregation.
Aggregation was correlated with cyclin-dependent kinases CDK4 and CDK6 and pathways characteristic of luminal cell subtypes, suggesting that blocking these pathways might promote more uniform distribution of tumor cells.
Predicting Efficacy with AI-Based Models
Based on their findings described above, the authors hypothesized that an AI-powered approach could identify biomarkers to predict ADC efficacy.
Using the available clinical data and spatial biomarker data from their biopsy samples, the authors first integrated response status, HR expression status, and computational pathology data consisting of cellular proportions and topological features. They used this integrated data asset to create an efficacy prediction model and assessed the performance of this model in SHR-A1811 training (AUC = 0.95) and testing (AUC = 0.86) data sets.
The high performance (AUC > 0.8) of this predictive model provides further support for the importance of spatial biomarkers, such as tumor cell aggregation and immune cell infiltration, in patient stratification.
A High-res AI-powered Approach to ADCs
At Nucleai, we’re redefining how translational and biomarker teams approach antibody-drug conjugate (ADC) research. Instead of relying on traditional methods like manual H-score assessments or simple presence/absence checks, our platform offers high-resolution, optical density-based IHC quantification and spatial analysis capabilities.
These tools provide deeper insights into how ADCs work, why resistance may occur, and how to better identify which patients will benefit most—while reducing the risk of harmful side effects. By integrating AI-powered spatial analytics with routine pathology data, we help teams make smarter, more targeted decisions in drug development.
Interested in learning how our platform can support your research goals? Schedule a demo to explore our latest innovations in AI-driven biomarker scoring and digital pathology.
References
-
Hurvitz, Sara A., et al. “Trastuzumab Deruxtecan versus Trastuzumab Emtansine in Patients with HER2-Positive Metastatic Breast Cancer: Updated Results from DESTINY-Breast03, a Randomised, Open-Label, Phase 3 Trial.” The Lancet, vol. 401, 2023, pp. 105–117. https://doi.org/10.1016/S0140-6736(22)02420-5.
-
Modi, Shanu, et al. “Trastuzumab Deruxtecan in Previously Treated HER2-Low Advanced Breast Cancer.” The New England Journal of Medicine, vol. 387, no. 1, 2022, pp. 9–20. https://doi.org/10.1056/NEJMoa2203690.
-
Bardia, Aditya, et al. “Trastuzumab Deruxtecan after Endocrine Therapy in Metastatic Breast Cancer.” The New England Journal of Medicine, vol. 391, no. 22, 2024, pp. 2110–2122. https://doi.org/10.1056/NEJMoa2407086.
-
Ma, Ding, et al. “Spatial Determinants of Antibody-Drug Conjugate SHR-A1811 Efficacy in Neoadjuvant Treatment for HER2-Positive Breast Cancer.” Cancer Cell, vol. 43, no. 6, 9 June 2025, pp. 1061–1075.e7. https://doi.org/10.1016/j.ccell.2025.03.017.
