Prediction of Acquired Taxane Resistance Using a Personalized Pathway-Based Machine Learning Method

Area of Science:

• Genomics
• Computational Biology
• Pharmacology

Background:

• Acquired taxane resistance (ATR) poses a significant challenge in cancer therapy.
• Accurate prediction of ATR is crucial for optimizing treatment strategies and improving patient outcomes.

Purpose of the Study:

• To develop and validate an individualized prediction model for automated detection of acquired taxane resistance (ATR).
• To enhance the generalizability and transferability of the predictive model across diverse datasets.

Main Methods:

• Utilized penalized regression and an individualized pathway score algorithm to construct the predictive model.
• Integrated multiple acquired taxane resistance (ATR) studies and employed robust cross-study validation for model development.
• Applied the model to public genomic cohorts for both acquired taxane resistance (ATR) and intrinsic taxane resistance (ITR).

Main Results:

• The ATR model achieved perfect internal cross-study validation (AUROC=1.000, AUPRC=1.000) and excellent performance on independent blind ATR cohorts (AUROC=0.940).
• Demonstrated superior accuracy for intrinsic taxane resistance (ITR) prediction (AUROC=0.70) compared to previous studies.
• Showcased high transferability on blind ATR cohorts (AUROC=0.69), indicating shared predictive features between intrinsic and acquired taxane resistance.

Conclusions:

• Successfully developed a personalized prediction model for acquired taxane resistance (ATR) using a multi-study approach.
• The model exhibits excellent accuracy, generalizability, and transferability, paving the way for automated ATR detection.
• Findings suggest potential common predictive features for both intrinsic and acquired taxane resistance, offering new avenues for therapeutic development.