Triple Negative Breast Cancer Markers in Liquid Biopsies Using Artificial Intelligence (TRICIA Study)

Rationale: The most aggressive form of breast cancer is triple negative breast cancer (TNBC), so called because these tumors do not express hormone receptors or HER2 receptor, and therefore have no readily targetable molecules. Chemotherapy is the only treatment, with chemoresistance signaling a very poor outcome even in early TNBC. The presence of residual tumor at surgery (non-pathological complete response or non-pCR) signals chemoresistance and poor prognosis, with about 30-40% of these patients dying of TNBC within the first 5 years after surgery. A recent clinical trial showed that the addition of further chemotherapy (Capecitabine) results in improved survival in these patients with non-pCR, although only about 15% of such patients do benefit. One of the most urgent unmet needs is to identify patients who will do well despite non-pCR (so as to avoid extra chemotherapy) and who will do poorly despite it, and also to identify factors of poor prognosis that may lead to novel therapeutic strategies in this group.

Current state of advancement of the technology: Until now, no biomarker except BRCA1/2 mutations has demonstrated clinical utility in the treatment of TNBC, likely due to the complex biology and heterogeneity of the disease. With the recent advances in Artificial Intelligence methodology, combining and integrating several layers of molecular data to predict outcome, until now challenging, becomes a reality. The hypothesize is that combining multi-dimensional data of tumor and plasma EVs can facilitate the development of prognostic and predictive signatures in this very aggressive disease.

Preliminary data: Thanks to our Q-CROC-03 biopsy driven clinical trial where tumor and plasma from patients with TNBC resistant to chemotherapy were collected. Whole exome seq data were translated to generate personalized circulating tumor DNA (ctDNA) assays. Our data shows a potential prognostic value to the detection of ctDNA after pre-operative chemotherapy. There is a collaboration established with Rodney Ouellette (ACRI) to isolate and profile extracellular vesicles (EVs) from plasma.

Objectives: The objective of the present study is to develop signatures of good and poor outcome as well of tumor response to chemotherapy in TNBCs by integrating multidimensional profiling of both tumor and liquid biopsies making use of Artificial Intelligence (AI) tools.

Experimental approach: EVs profiling from plasma collected in the Q-CROC-03 trial and the JGH biobank (prior, during and after chemotherapy treatment) will be performed. Profiling will include Whole Genome Sequencing (GWS), proteomics, transcriptomics and miRNA analysis. In collaboration with our industrial partner, My Intelligent Machines (MIMs), experts in bioinformatics and AI, machine-learning algorithms will be developed to integrate OMICs data from resistant tumors with matched plasma EVs data and generate a tumor/plasma signature associated with poor outcome. In parallel, in collaboration with the EXACTIS Innovation Network, patients recruitment, collection of residual tumors post chemotherapy and matched serial plasma samples during capecitabine treatment after surgery to perform the validation of the signature identified, the tumor/EV signature will be associated with patient survival.