BrcaDx

BrcaDx classifies breast cancer versus normal tissue using gene expression biomarkers to provide high-confidence diagnostic discrimination for breast cancer detection.

Key Features:

• Optimized Biomarker Identification: Uses transcriptomic profiles from public-domain datasets to identify nine progression-significant genes: NEK2, PKMYT1, MMP11, CPA1, COL10A1, HSD17B13, CA4, MYOC, and LYVE1.
• Advanced Machine Learning Techniques: Employs feature selection, principal components analysis (PCA), and k-means clustering to train a model that separates cancerous and normal samples based on biomarker expression.
• High Accuracy and Validation: Reports 99.5% accuracy on independent test datasets and 95.5% balanced accuracy on out-of-domain external validation datasets.

Scientific Applications:

• Early Detection: Provides a molecular diagnostic approach that can discriminate cancerous from normal breast tissue using minimal biomarker sets for potential early detection.
• Dimensionality Reduction: Reduces gene expression data complexity by focusing on a small set of progression-significant biomarkers to enhance interpretability and downstream analysis.

Methodology:

Transcriptomic profiles from breast cancer samples were screened using stage-informed models to identify progression-significant genes; feature selection, PCA, and k-means clustering were applied to train a model that distinguishes "cancer" from "normal" based on selected biomarker expression; the model was validated on independent and out-of-domain external datasets, yielding the reported accuracy metrics.