Dissecting the Genome for Drug Response Prediction

Area of Science:

• Computational biology
• Genomics
• Pharmacology

Background:

• Precision medicine requires accurate prediction of cancer cell line drug sensitivity.
• Omics and pharmacogenomics data are available for over 1000 cancer cell lines.
• Machine learning (ML) and deep learning (DL) models are increasingly used for drug sensitivity prediction.

Purpose of the Study:

• To review recent advances in computational methods for cancer cell line drug sensitivity prediction.
• To explore methods that integrate diverse data sources, including drug chemical information and prior feature selection.
• To discuss emerging approaches for predicting clinical drug response from cell line data.

Main Methods:

• Review of computational methods published in the last two years.
• Analysis of methods applied to public datasets: COSMIC, Cancer Cell Line Encyclopedia (CCLE), and National Cancer Institute-60 (NCI-60).
• Focus on ML/DL algorithms incorporating omics, pharmacogenomics, and drug chemical data.

Main Results:

• Several representative computational methods have been developed to enhance drug sensitivity prediction.
• Methods incorporating drug chemical information or a priori feature selection show improved predictive performance.
• Recent advancements aim to bridge the gap between cell line drug sensitivity and clinical patient response.

Conclusions:

• Integrating multi-omics data and drug features significantly improves cancer drug sensitivity prediction.
• Advanced computational methods are key to realizing precision medicine's potential.
• Translating cell line findings to clinical outcomes remains an active area of research.