Related Experiment Video
Updated: Dec 6, 2025
03:08
Using Human Differentially Expressed Gene Lists to Perform Downstream Pathway Enrichment Analysis and Target Prioritization
Published on: October 3, 2025
780
Pathway-Based Drug Response Prediction Using Similarity Identification in Gene Expression
Seyed Ali Madani Tonekaboni1,2, Gangesh Beri1, Benjamin Haibe-Kains1,2,3,4,5
1Princess Margaret Cancer Centre, Toronto, ON, Canada.
Frontiers in Genetics
|October 9, 2020
Summary
Identifying biomarkers for HER2+ breast cancer drug response is crucial. Pathway activity, not individual genes, accurately predicts response to lapatinib and trastuzumab therapies.
Area of Science:
- Oncology
- Genomics
- Bioinformatics
Background:
- Targeted therapies like lapatinib and trastuzumab improve outcomes for HER2+ breast cancer patients.
- However, a significant portion of patients do not respond to these treatments, necessitating better predictive biomarkers.
Purpose of the Study:
- To identify reliable biomarkers for predicting patient response to lapatinib and trastuzumab in HER2+ breast cancer.
- To compare the predictive power of gene expression profiles, pathway activities, and machine learning models.
Main Methods:
- Analyzed gene expression profiles from the NeoALTTO clinical trial.
- Assessed univariate gene expression biomarkers and pathway activities using the transcriptional similarity coefficient (TSC).
- Developed and evaluated machine learning models (logistic regression, naive Bayes, random forest, k-NN, SVM) for response prediction.
Main Results:
- Univariate gene expression biomarkers showed weak predictive value.
- Pathway activities estimated via TSC demonstrated high predictive accuracy for therapy response (concordance index >0.8, p < 0.05).
- Pathway-based approach outperformed various machine learning models in predicting drug response.
Conclusions:
- Transcriptional similarity of biological pathways is a robust predictor of response to lapatinib and trastuzumab in HER2+ breast cancer.
- This pathway-centric approach offers a more effective strategy for personalized therapy selection compared to individual gene biomarkers or standard machine learning models.
