I-Biomarker CaDx: A circulating miRNA-based multi-cancer detection tool with explainable AI for breast cancer.

Abstract

5 Background: Breast cancer is requires early and accurate detection for effective treatment and improved patient outcomes. Current methods, such as mammography and biopsy, have limitations in accuracy, invasiveness, and accessibility. Circulating microRNAs (miRNAs) are promising non-invasive biomarkers for cancer detection, including breast cancer. i-Biomarker CaDx is a technology-agnostic, multi-cancer early detection and diagnosis test, covering 13 cancer types with 99-100% accuracy. This study evaluates i-Biomarker CaDx performance for breast cancer using a publicly available dataset and assesses explainable artificial intelligence (XAI) interpretations. Methods: The dataset (GEO: GSE73002) contains 1,280 breast cancer patients and an equal number of age-matched healthy individuals. Circulating miRNA profiles were measured using microarray technology. Multiple classification algorithms, including an ensemble of decision trees, neural networks, and support vector machines, were employed for feature selection, classification, and XAI. Hyperparameter optimization was performed, and the best classifiers were integrated with weights proportional to their performance into a final model. The performance of i-Biomarker CaDx was evaluated using cross-validation and independent test sets, while the XAI component provided cohort-level and individual-level explanations of the miRNA alterations associated with the diagnostic outcome. Results: i-Biomarker CaDx demonstrated remarkable diagnostic performance in detecting breast cancer, achieving 100% accuracy in the studied dataset. The sensitivity, specificity, positive predictive value, and negative predictive value of i-Biomarker CaDx surpassed those of traditional diagnostic methods such as mammography and biopsy. The ensemble classifier, consisting of decision trees, neural networks, and support vector machines, effectively captured the complex interactions among miRNAs. The XAI component of i-Biomarker CaDx provided valuable insights into the biological relevance of miRNA alterations in breast cancer diagnosis. Cohort-level and individual-level explanations elucidated the miRNA patterns and their associations with breast cancer, enhancing the understanding of the underlying molecular mechanisms. Conclusions: i-Biomarker CaDx, a technology-agnostic multi-cancer early detection and diagnosis tool, has shown exceptional accuracy in breast cancer detection, outperforming traditional diagnostic methods. The integration of explainable artificial intelligence (XAI) facilitates a deeper understanding of miRNA alterations and their associations with breast cancer, contributing to improved interpretation at both the cohort and individual levels. The results of this study demonstrate the potential of i-Biomarker CaDx as a powerful, non-invasive diagnostic tool for breast cancer.