Multiomics Analysis of Gene Expression and Drug Sensitivity Relationship in Breast Cancer Subtypes

Read full text Summarise with VIRAT
Authors: Adem SONUVAR, Esra Tokur Sonuvar, Uğur Bilge
Cited by: 0
References: 28

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Breast cancer is a complex disease exhibiting distinct clinical behaviors due to its molecular heterogeneity. In this study, 30 breast cancer cell lines were categorized into two main analysis groups: 'Hormone/HER2 Positive' (n=12) and 'TNBC' (triple-negative breast cancer). Gene expression profiles (RNA-seq), proteomic data (RPPA), and drug sensitivity to seven different kinase inhibitors were analyzed using the E-MTAB-4801 dataset. t-SNE and hierarchical clustering demonstrated a separation based on the overall expression profiles of these two groups DEG (edgeR; FDR1) and enrichment analyses (GO/KEGG via clusterProfiler; Hallmark via msigdbr/fgsea) identified significant activation of estrogen response and metabolic pathways in the Hormone/HER2 Positive group, whereas immune response and epithelial-mesenchymal transition pathways were active in the TNBC group. Machine learning methods identified potential biomarker candidates (FAM176A, CACNG1, GPR77) crucial for discriminating between the two groups. Model performance was evaluated using nested cross-validation (five outer folds). Across outer folds, LASSO achieved mean Accuracy=0.833, F1=0.865, Sensitivity=0.889, Specificity=0.750; Random Forest achieved Accuracy=0.800, F1=0.833, Sensitivity=0.833, Specificity=0.750; and the Decision Tree showed more variable performance with mean Accuracy=0.700. Furthermore, analyses using RPPA data confirmed that the 4E-BP1 protein expression level is an important determinant of sensitivity to mTOR inhibitors. However, findings are based on 30 breast cancer cell lines, which limits statistical power and increases the risk of overfitting, and no independent clinical cohort was available for external validation; therefore, the identified biomarkers should be considered candidate biomarkers requiring validation in patient cohorts. Accordingly, the results are intended to support research-stage subtype stratification and prioritization of biomarker candidates rather than immediate clinical decision-making or drug selection providing a prioritized shortlist of candidates for follow-up validation in patient tumor cohorts.

Fields of Study & Subjects

Breast Cancer Treatment Studies Gene expression and cancer classification Bioinformatics and Genomic Networks

Publication details

Journal / source
AJIT-e Academic Journal of Information Technology
Volume
17
Issue
1
Pages
26–47
Publication date
28 February 2026
Language
EN