Deep learning for drug response prediction in cancer

Area of Science:

• Computational biology
• Pharmacogenomics
• Artificial intelligence in oncology

Background:

• Predicting anti-cancer treatment sensitivity is crucial for precision medicine.
• Machine learning (ML) models can predict drug response using high-throughput screening data.
• Deep learning (DL), a subset of ML, offers potential for complex drug response modeling.

Purpose of the Study:

• To critically review recent studies using DL for predicting drug response in cancer cell lines.
• To provide an overview of DL architectures applied in this domain.
• To highlight data resources and discuss limitations and future improvements.

Main Methods:

• Review of published literature on DL for drug response prediction in cancer.
• Description of common DL architectures (e.g., convolutional neural networks, recurrent neural networks).
• Identification and curation of publicly available drug screening datasets.

Main Results:

• Emerging studies demonstrate promising results for DL in predicting cancer cell line drug sensitivity.
• DL models show potential for integrating diverse biological and chemical data.
• The field is rapidly evolving with increasing research interest.

Conclusions:

• DL is a promising approach for enhancing the prediction of anti-cancer drug response.
• Further research is needed to address limitations and optimize DL models.
• Standardized data resources and validation are essential for clinical translation.