Machine learning-assisted prognostication based on genomic expression in the tumour microenvironment of estrogen receptor positive and HER2 negative breast cancer

Abstract

Background: Stroma in the tumor microenvironment (TME) is known to impact prognosis and responses to therapy. Few mathematical models exist to prognosticate patients (pts), based on mRNA expressivity in the TME. Method(s): Clinical outcomes data and mRNA-seq of 246 pts with stage 2 estrogen receptor (ER) positive (+) and HER2 negative (-) breast cancer were obtained from TCGA. 26 gene groups composed of 191 genes* enriched in cellular and non-cellular elements of TME, mutational burden (MB), and clinical data were analyzed by Kaplan-Meier (KM) analysis and multivariate nonlinear regression assisted by machine learning to achieve confined optimization with model-data minimization among multiple distribution functions. *Due to character limit, more details about these genes will be shown at actual presentation. Result(s): Prognostication was modeled with higher risk score (RS) representing worse prognosis in stage 2 ER+HER2-breast cancer. Six genes (C15orf53, PDGFB, IL10, HS3ST2, GPNMB, PADI4) and seven genes (FCRL3, IFNGR2, ICAM2, CXCR4, HLADMB, LGMN, ICOSLG) out of 191 genes associated with poor prognosis were identified (p<0.05 and 0.05 <0.001, HR=3.762 (95% CI 2.914-4.939), confirming the validity of RS modeling. Analysis of immune profiles in high and low RS groups shows that expression of genes associated with immunosuppressive factors, T-helper 2 cells, macrophages, neutrophils, co-inhibitory factors of T-cells, and antigen presenting cells are higher in high RS group (p<0.05). MB did not contribute to survival. Conclusion(s): Machine learning-assisted mathematical modeling of RS and gene analysis identified TME-related genes and gene groups that are strongly associated with worse prognosis in stage 2 ER+HER2-breast cancer. RS could potentially prognosticate pts in the clinic with available genomic profiles.