Developing a gene expression classifier for breast cancer diagnosis

Breast cancer (BC) is the most common type of cancer in women worldwide. Solid tumors are complex structures composed of many cell types and extracellular matrix components. Understanding solid tumors is crucial for developing effective treatments. This study aimed to develop a gene expression classifier to predict BC with high accuracy. The study first identified the most important genes for cancer through differential expression analysis (DEA) between breast cancer and adjacent normal breast samples. The R package STRINGdb was then used to create a protein-protein interaction network (PPI) to examine upregulated genes and find clusters. Enrichment analyses were performed to identify overrepresented biological functions and pathways. A logistic regression prediction model was developed using a breast cancer dataset from TCGA and evaluated using discrimination and calibration measures. BUB1 expression in breast cancer was also investigated using quantitative analysis. Two significant clusters were identified, with cell cycle checkpoints and M phase key pathways in one cluster and extracellular matrix organization in the other. A prediction model using the hub gene set (COMP, FN1, SDC1, BUB1, TTK, and NUSAP1) showed high sensitivity (97.2%) and specificity (96.1%), and an AUC of 0.994. Three hub genes (COMP, FN1, and SDC1) were identified through the PPI network, strongly linked to extracellular matrix organization (BUB1, TTK, and NUSAP1) as hub genes involved in M phase and cell cycle checkpoints. Overall, the study identified hub pathways and genes that accurately distinguish between cancer and normal samples, presenting promising new possibilities for early cancer detection and improved BC therapy.