ADRML: anticancer drug response prediction using manifold learning.

Area of Science:

• Computational biology
• Pharmacogenomics
• Bioinformatics

Background:

• Precision medicine necessitates personalized treatment strategies based on individual patient data.
• Large-scale drug and cell line datasets enable the development of computational models for predicting anticancer drug response.

Purpose of the Study:

• To propose ADRML (Anticancer Drug Response Prediction using Manifold Learning), a model for accurate anticancer drug response prediction.
• To integrate cell line and drug information using manifold learning for enhanced therapeutic predictions.

Main Methods:

• ADRML maps the drug response matrix into lower-rank spaces to derive new insights into cell lines and drugs.
• Low-rank features are utilized to compute drug response for novel cell line-drug pairs.

Main Results:

• ADRML demonstrates accurate and robust predictions across diverse cell lines and drug data.
• The model's predictions correlate with pathway activity scores, offering insights into drug mechanisms.
• Case studies validate ADRML's predictions for novel cell line-drug pairs using existing literature.

Conclusions:

• ADRML effectively predicts and imputes anticancer drug response, supporting precision medicine initiatives.
• The model provides valuable perspectives on cell line-drug interactions and potential therapeutic strategies.