Drug response prediction by inferring pathway-response associations with kernelized Bayesian matrix factorization.

Area of Science:

• Computational biology
• Genomics
• Pharmacogenomics

Background:

• Personalized medicine aims to predict drug responses using molecular data.
• Cancer's molecular heterogeneity poses challenges for predicting treatment efficacy.
• Accurate drug response prediction is crucial for tailoring therapies to individual patients.

Purpose of the Study:

• To develop a novel computational method for predicting drug responses in unseen samples.
• To integrate multi-omics data and pathway information for improved prediction accuracy.
• To enhance the prediction of drug responses in oncology.

Main Methods:

• A novel multi-task matrix factorization formulation was developed.
• The method extends kernelized Bayesian matrix factorization (KBMF) with component-wise multiple kernel learning.
• Exploits pathway information for biologically meaningful learning of drug response associations.

Main Results:

• The proposed method quantitatively outperforms state-of-the-art approaches on cancer datasets.
• Predictions were validated using laboratory experiments on a cancer cell line panel.
• The model successfully inferred pathway-drug response associations for EGFR and MEK inhibitors.

Conclusions:

• The developed method offers a powerful tool for computational personalized medicine.
• It enables more accurate prediction of drug responses, aiding in therapy selection.
• The approach facilitates the elucidation of drug action mechanisms through pathway analysis.