Alexios Matikas

The ongoing revolution in the treatment of HER2-positive breast cancer has led to the rapid approval of multiple drugs with broad efficacy, potential for toxicity and high associated costs for all-comer populations with no attempt for biomarker discovery. In the ongoing academic, international, randomized ARIADNE trial (NCT05900206) which the applicant leads as co-PI we compare preoperative trastuzumab deruxtecan and standard of care chemotherapy combined with anti-HER2 antibodies (TCHP), and we allocate patients to treatment according to the underlying biology of their tumors by using PAM50 subtypes. With this project, we aim to leverage multi-omic analyses of baseline samples (whole exome and RNA-sequencing, mass spectrometry proteomics) and extraction of pathomics and radiomics features from whole slide images and MRI images, respectively. We will then perform machine learning-based multimodal integration of the best predictive, not simply prognostic, features to construct a predictive model for patient selection for treatment with antibody-drug conjugate or chemotherapy and dual HER2-blockade combination. The uniqueness of the clinical trial, revolutionary efficacy of trastuzumab deruxtecan, comprehensive methodology, and rigorous external validation underscore both the scientific novelty and the potential clinical significance of the proposed project.

Breast cancer is the most common form of cancer in women - one in 8 women will develop breast cancer in their lifetime. A large number of studies of tens of thousands of women with early breast cancer have shown that chemotherapy after surgical resection of the tumor improves survival in these patients. One such study was the Swedish PANTHER study. Chemotherapy can also be given before resection with equivalent effect, as in the Swedish PREDIX Lum B study and the international EORTC "p53" study. Lack of reliable markers sometimes leads to undertreatment and more often to overtreatment, which means more side effects and increased costs. The project concerns the analysis of tumor material already taken from patients who participated in the PANTHER, PREDIX Lum B and EORTC studies and focuses on hormone-sensitive breast cancer, which is considered a less aggressive variant. However, these patients have a long-standing and increasing risk of cancer recurrence and benefit greatly from chemotherapy, while some are exposed to unnecessary side effects. Gene expression, clinical factors and the immune microenvironment will be analyzed and the factors with the greatest predictive/prognostic value will then be integrated into a model expected to have a much higher predictive capacity compared to models from single sources. With this project, we intend to create a new prognostic and predictive tool that can help us select the right treatment for the right patient with hormone-sensitive early breast cancer - 70% of all breast cancer cases. In other words, we will create a tool that can help us limit chemotherapy to only those breast cancer patients who need it most. In the long term, this project is expected to lead to clinically useful markers to be able to select, on an individual level, suitable candidates who are treated with optimized treatments: better effect, fewer side effects and reduced costs are the project's goals.