On April 13, 2026 Genialis, the therapeutic intelligence company, reported it will present the first results from a new AI-based algorithm designed to distinguish between patients more likely to benefit from treatment with trastuzumab deruxtecan (Enhertu, T-DXd). Built using the Genialis Supermodel, and evaluated on real-world clinical data from Tempus, including a cohort of 90 breast cancer patients with HER2-positive, HER2-low and HER2-ultra-low disease, the model demonstrates statistically significant discriminatory performance that outperforms standard HER2 diagnostics. The full poster will be presented at the American Association for Cancer Research (AACR) (Free AACR Whitepaper) Annual Meeting 2026 (April 17–22; San Diego).
Schedule your 30 min Free 1stOncology Demo!
Discover why more than 1,500 members use 1stOncology™ to excel in:
Early/Late Stage Pipeline Development - Target Scouting - Clinical Biomarkers - Indication Selection & Expansion - BD&L Contacts - Conference Reports - Combinatorial Drug Settings - Companion Diagnostics - Drug Repositioning - First-in-class Analysis - Competitive Analysis - Deals & Licensing
Schedule Your 30 min Free Demo!
Antibody-drug conjugates (ADCs) are among the fastest-growing drug classes in oncology, yet patient selection still relies on diagnostics that measure target expression without capturing the tumor biology that determines response. For HER2-directed ADCs like Enhertu, that gap spans the full HER2 spectrum: clinical benefit varies substantially across HER2-measured disease status, and IHC (Immunohistochemistry) or FISH (Fluorescence In Situ Hybridization) testing cannot reliably identify who will respond.
Genialis used the Genialis Supermodel, a large molecular foundation model trained on billions of RNA-seq-derived data points, to develop a predictive model for T-DXd response. The Supermodel translates tumor gene expression into hundreds of biomodules, or machine learning-derived representations of biological processes. By incorporating biomodules relevant to ADC mechanisms, including targeting, internalization, and payload activity, the model captures the underlying biology associated with treatment response.
The model was developed and evaluated in a real-world clinical cohort of T-DXd-treated breast cancer patients from the Tempus multimodal database. Using real-world clinical benefit duration (CBD) as the survival endpoint, the model achieved statistically significant predictive discrimination: hazard ratio 2.22 (95 percent CI 1.14–4.35), p = 0.020. Patients predicted as likely to benefit had a median duration of treatment of 345 days, compared to 245 days for those patients unlikely to benefit, a 41 percent difference. No prognostic signal was observed in control cohorts (HR ≈ 1.0), confirming that the model captures treatment-specific benefit rather than general prognosis.
"Better biomarkers for HER2-directed ADCs are urgently needed as current diagnostics were designed to measure receptor expression, not the full biology of drug response," said Mark Uhlik, PhD, Chief Scientific Officer of Genialis. "Enhertu response involves multiple biological processes beyond HER2 expression, including internalization, payload activity, and the tumor’s damage response. These aren’t captured by IHC. The Supermodel is designed to represent those processes using comprehensive RNA data, which helps explain the signal we’re presenting at AACR (Free AACR Whitepaper)."
The model’s strongest predictive features are associated with ADC biology rather than HER2 expression alone, including topoisomerase payload activity, DNA damage response, hypoxia, and tumor stress pathways, illustrating why receptor-level diagnostics cannot capture the full complexity of ADC response.
Improving patient selection remains a major challenge in cancer drug development, especially as ADC pipelines continue to expand. With more than 1,000 ADC programs in clinical development, there is a growing demand for biomarker strategies that can keep pace with the diversity of these therapies. The Genialis Supermodel’s modular architecture independently captures each biological step of ADC mechanism, making it applicable to any antibody-payload combination. Genialis is already applying this framework to multiple ADC programs in collaboration with pharmaceutical and biotech partners.
"HER2 IHC tells you what a tumor expresses. The Supermodel tells you the biology present in a tumor and predicts what will happen when you treat it. That is a fundamentally different kind of information, and it is what drug developers need to make smarter decisions earlier in ADC development," said Rafael Rosengarten, PhD, Chief Executive Officer of Genialis. "This Enhertu predictor is another example of the Supermodel’s broad application across the ADC pipeline, and we are rapidly extending this work with our pharma and biotech partners."
Poster Presentation Details:
Title: "An RNA-Based Survival Model Predicting Real-World Response to Trastuzumab Deruxtecan"
Meeting: AACR (Free AACR Whitepaper) Annual Meeting 2026
Date: April 22, 2026 | 9:00 AM– 12:00 PM
Location: San Diego Convention Center — Poster Section 3, Board 27
Poster Number: 6883
For more information on Genialis and the Supermodel platform, including krasID and Expressions, visit www.genialis.com.
(Press release, Genialis, APR 13, 2026, View Source [SID1234664335])