# AI quantification of HER2-low and HER2-ultralow

Source: https://onco.cc/ideas/idea-ai-her2-low-scoring/  
OnCo record `idea-ai-her2-low-scoring` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Pathologists disagree about faint HER2 staining, yet that decision unlocks Enhertu. Let a validated algorithm do the counting.

## Summary

Pathologists disagree about faint HER2 staining, yet that call decides who receives trastuzumab deruxtecan, so this idea hands the counting to a validated algorithm. A continuous AI-derived HER2 membrane score is expected to predict T-DXd benefit better than pathologist-assigned IHC categories and to reclassify some IHC 0 tumours as eligible, because benefit depends on a continuous delivery threshold that binned human scoring discards. The test is a retrospective analysis of DESTINY-Breast04 and DESTINY-Breast06 slides with an AI scorer, then prospective companion-diagnostic validation. At early-clinical maturity it addresses the bottlenecks Biomarkers are not validated or standardised and AI that is built but not validated or deployed.

## Fields

- Kind: Idea
- Last checked: 2026-09-04
- Hypothesis: A continuous AI-derived HER2 membrane score predicts T-DXd benefit better than pathologist-assigned IHC categories, and reclassifies a meaningful fraction of IHC 0 tumours as eligible.
- Rationale: T-DXd benefit depends on a delivery threshold, which is continuous; discretised human scoring loses information.
- Proposed test: Run a retrospective analysis of DESTINY-Breast04/06 slides with an AI scorer, then prospective companion-diagnostic validation.
- Maturity: early-clinical

## Sources

- DESTINY-Breast04: trastuzumab deruxtecan works in HER2-low breast cancer, creating a new treatable group (New England Journal of Medicine 2022): https://doi.org/10.1056/NEJMoa2203690

## Connected records

- pairings: [HER2-low scoring → T-DXd](https://onco.cc/pairings/her2-low-to-tdxd/)
- cancers: [HR-positive / HER2-negative breast cancer](https://onco.cc/cancers/breast-hr-positive/), [Triple-negative breast cancer (TNBC)](https://onco.cc/cancers/tnbc/)
- technologies: [Digital pathology & AI](https://onco.cc/technologies/digital-pathology-ai/), [Histopathology & immunohistochemistry](https://onco.cc/technologies/histopathology-ihc/)
- drugs: [Trastuzumab deruxtecan](https://onco.cc/drugs/trastuzumab-deruxtecan/)
- terms: [HER2-low and HER2-ultralow](https://onco.cc/terms/her2-low/), [Immunohistochemistry (IHC)](https://onco.cc/terms/ihc/)
- trials: [DESTINY-Breast04](https://onco.cc/trials/destiny-breast04/), [DESTINY-Breast06](https://onco.cc/trials/destiny-breast06/), [Trastuzumab Deruxtecan (T-DXd) in Patients Who Have Hormone Receptor-negative and Hormone Receptor-positive HER2-low or HER2 IHC 0 Metastatic Breast Cancer](https://onco.cc/trials/nct05950945/)
- key papers: [Antibody-drug conjugates in metastatic triple negative breast cancer: a spotlight on sacituzumab govitecan, ladiratuzumab vedotin, and trastuzumab deruxtecan](https://onco.cc/key-papers/paper-trastuzumab-deruxtecan-tnbc-expert-opin-biol-ther-2021/), [Neoadjuvant trastuzumab deruxtecan alone or followed by paclitaxel, trastuzumab, and pertuzumab for high-risk HER2-positive early breast cancer (DESTINY-Breast11): a randomised, open-label, multicentre, phase III trial](https://onco.cc/key-papers/paper-trastuzumab-deruxtecan-breast-hr-positive-ann-oncol-2026/), [Sacituzumab govitecan and trastuzumab deruxtecan: two new antibody-drug conjugates in the breast cancer treatment landscape](https://onco.cc/key-papers/paper-trastuzumab-deruxtecan-tnbc-esmo-open-2021/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Biomarkers are not validated or standardised](https://onco.cc/bottlenecks/b-biomarker-validation/)
- ideas: [A fund for prospective validation of academic biomarkers and companion diagnostics](https://onco.cc/ideas/idea-fund-biomarker-validation-fund/), [A standard evaluation pathway for AI-assisted pathology, from reader study to deployment](https://onco.cc/ideas/idea-data-ai-pathology-evaluation-standard/), [AI-first reading for high-volume common cancer diagnoses, pathologist for the exceptions](https://onco.cc/ideas/idea-acc-ai-first-pathology-common-cases/), [Calibrated reference slides so every lab scores HER2-low the same way](https://onco.cc/ideas/idea-tr2-her2-low-reference-materials/), [Version control and locked reference sets for AI algorithms used as companion diagnostics](https://onco.cc/ideas/idea-tr2-ai-cdx-change-control/)

---
JSON: https://onco.cc/api/v1/entities/idea-ai-her2-low-scoring.json