Gamma distribution based predicting model for breast cancer drug response based on multi-layer feature selection.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Precision medicine in oncology relies on predicting drug response from patient genomic data.
• Existing machine learning models often provide point predictions, lacking information on result distribution and reliability.
• Accurate prediction of drug sensitivity is crucial for effective cancer treatment strategies.

Purpose of the Study:

• To develop and evaluate novel machine learning approaches for predicting anticancer drug response from genomic data.
• To address the limitations of point prediction by incorporating prediction intervals for enhanced reliability.
• To improve clinical decision support for cancer patients through more accurate and interpretable drug response predictions.

Main Methods:

• Proposed a three-layer feature selection combined with a gamma distribution-based Generalized Linear Model (GLM).
• Developed a two-layer feature selection combined with an Artificial Neural Network (ANN).
• Applied and validated these methods on the Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug Sensitivity in Cancer (GDSC) datasets using ten-fold cross-validation.

Main Results:

• Achieved higher accuracy in anticancer drug response prediction compared to existing methods, with R-squared of 0.87 and RMSE of 0.53.
• Demonstrated the significance of confidence intervals for assessing prediction reliability.
• Correlation analysis confirmed the effectiveness of the selected features from the multi-layer selection process.

Conclusions:

• The proposed GLM and ANN methods offer improved accuracy and reliability for predicting anticancer drug response.
• Incorporating confidence intervals enhances the clinical utility of predictive models in personalized medicine.
• The feature selection strategies effectively identify relevant genomic markers for drug response prediction.