Abstract:Background: Anti-HER2 targeted therapy has achieved a series of breakthroughs. However, the current treatment strategy regarding HER2-positive breast cancer remains indiscriminate and lacks specificity, which limits the further improvement of overall treatment response and may lead to overtreatment and extra cost for some patients. Our study aims to reveal the molecular heterogeneity of HER2-positive breast cancer to guide a more precise treatment. Patients and methods: We selected HER2-positive breast cancer patients treated at Fudan University Shanghai Cancer Center between 2013 and 2014 and conducted genomic, transcriptomic, proteomic and metabolomic profiling. We then applied a non-negative matrix factorization algorithm on transcriptomic data to obtain an unsupervised classification. And we further studied the correlation between subtypes and corresponding treatment strategies in multiple cohorts of adjuvant and neoadjuvant therapy. For clinical accessibility, we developed convolutional neural network models through deep learning algorithm based on digital pathology to identify different subtypes. Additionally, we explored novel treatment strategies using the patient-derived organoids (PDOs) models. Results: We established a novel multiomics cohort of 180 HER2- breast cancer patients and classified them into four clinically significant molecular subtypes: (1) A classical HER2-enriched (HER2-CLA, N=51) subtype characterized by strong ERBB2 signaling and remarkable sensitivity to anti-HER2-targeted therapy (pathologic complete response with dual-targeted therapy: 93%). (2) an immunomodulatory (HER2-IM, N=36) subtype characterized by an immune-activated microenvironment and excellent prognosis with current treatment (no relapse in 97% of patients with a median follow-up of 86 months). Tumors of this subtype were therefore candidates for de-escalatory treatment. (3) A luminal-like (HER2-LUM, N=55) subtype distinguished by activated estrogen receptor signaling and (4) a basal/mesenchymal-like (HER2-BM, N=38) subtype enriched in activated receptor tyrosine kinase pathways. HER2-LUM and HER2- BM showed limited benefit from anti-HER2 therapy, and thus, add-on therapies might be needed. The overall area under the curve (AUC) of the convolutional neural network model based on digital pathology for identifying different subtypes is 0.77. In the exploration of novel treatment strategies, we found in the PDO model that the HER2-LUM subtype is more sensitive to a treatment regimen combining standard (chemotherapy and targeted therapy) with subsequent endocrine therapy and CDK4/6 inhibitors compared to other subtypes. Additionally, the HER2-BM subtype demonstrated greater sensitivity to treatment with a combination of EGFR inhibitors, PDGFR inhibitors or VEGFR inhibitors. Conclusion: We uncovered a high degree of molecular heterogeneity in HER2-positive breast cancer and illustrated its impact on treatment response. More precise treatment can be given according to the characteristics of different subtypes, which may achieve good efficacy and simultaneously reduce overtreatment and extra cost. The comprehensive profiling of HER2-positive breast cancers could also serve as an important resource for further exploration. Key Words: HER2-positive breast cancer cohort; molecular classification; targeted therapy; precision treatment; de-escalatory treatment. Citation Format: Yu-Wei Li, Ding Ma, Xiang-Rong Wu, Lei-Jie Dai, Shen Zhao, Yu-Zheng Xu, Xi Jin, Xiao Yi, Ying Wang, Cai-Jin Lin, Yi-Fan Zhou, Tong Fu, Wen-Tao Yang, Ming Li, Hong Lv, Yi-Zhou Jiang, Zhi-Ming Shao. Multiomics profiling and molecular classification refine precision treatment strategies for HER2-positive breast cancer [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl) nr PS09-09.

Themis: advancing precision oncology through comprehensive molecular subtyping and optimization

Reassessment of Reliability and Reproducibility for Triple-Negative Breast Cancer Subtyping

Identification of Molecular Subtypes for Integrated Multi-Omics Analysis for Use in Guiding Precision Medicine in Hepatocellular Carcinoma.

Molecular Subtyping of Hepatocellular Carcinoma: A Step Toward Precision Medicine

Molecular Characterization and Classification of HER2-Positive Breast Cancer Inform Tailored Therapeutic Strategies

Molecular Subtyping of Cancer Based on Distinguishing Co-Expression Modules and Machine Learning

Generalizable transcriptome-based tumor malignant level evaluation and molecular subtyping towards precision oncology

Evaluation and comparison of multi-omics data integration methods for cancer subtyping

Subtype-Former: a deep learning approach for cancer subtype discovery with multi-omics data

Decoding transcriptomic intra-tumour heterogeneity to guide personalised medicine in ovarian cancer

Abstract PS09-09: Multiomics profiling and molecular classification refine precision treatment strategies for HER2-positive breast cancer

Abstract PO1-15-07: Subtyping-directed precision treatment refines traditional one-size-fits-all therapy in HR+/HER2- breast cancer: a sub-study of the MULAN umbrella trial

Uncovering the Subtype-Specific Molecular Characteristics of Breast Cancer by Multiomics Analysis of Prognosis-Associated Genes, Driver Genes, Signaling Pathways, and Immune Activity

Integrative Multiomics Analyses Identify Molecular Subtypes of Head and Neck Squamous Cell Carcinoma with Distinct Therapeutic Vulnerabilities

Therapeutic Response-Based Reclassification of Multiple Tumor Subtypes Reveals Intrinsic Molecular Concordance of Therapy Across Histologically Disparate Cancers

Molecular Subtyping of Triple-Negative Breast Cancers by Immunohistochemistry: Molecular Basis and Clinical Relevance

An overview of multiomics: a powerful tool applied in cancer molecular subtyping for cancer therapy

Identification of Breast Cancer Subtypes by Integrating Genomic Analysis with the Immune Microenvironment

Integrative Multi-Omics Analyses Identify Molecular Subtypes of Head and Neck Squamous Cell Carcinoma with Distinct Therapeutic Vulnerabilities

ProTICS Reveals Prognostic Impact of Tumor Infiltrating Immune Cells in Different Molecular Subtypes.

92-Gene Molecular Profiling in Identification of Cancer Origin: a Retrospective Study in Chinese Population and Performance Within Different Subgroups.