Clinical Drug Response Prediction by Using a Lq Penalized Network-Constrained Logistic Regression Method

Area of Science:

• Genomics
• Pharmacogenomics
• Computational Biology

Background:

• Personalized medicine relies on understanding the link between patient genotypes and drug responses.
• Current predictive models for genotype-drug response relationships require improvement.

Purpose of the Study:

• To develop a novel computational method to enhance the prediction of drug response based on genetic profiles.
• To integrate gene expression data with biological network knowledge for improved predictive accuracy.

Main Methods:

• Introduction of the Lq penalized network-constrained logistic regression (Lq-NLR) method.
• Integration of gene expression data and biological network information with a robust penalty function.
• Development of response prediction models for erlotinib and sorafenib using cell line data.

Main Results:

• The Lq-NLR method demonstrated high effectiveness in predicting drug response.
• Achieved a significant correlation of 0.841 between in vitro and in vivo drug response predictions.
• Successfully applied prediction models to develop a personalized medicine approach based on gene expression profiles.

Conclusions:

• The proposed Lq-NLR method significantly outperforms existing approaches.
• This method provides a more accurate reflection of genotype-phenotype relationships in drug response.
• Enables more precise personalized cancer treatment strategies.