Decoding Drug Response With Structurized Gridding Map-Based Cell Representation.

Area of Science:

• Pharmacology
• Computational Biology
• Genomics

Background:

• Understanding cell-line drug response is vital for effective drug development and overcoming resistance.
• Current methods, like single-gene analysis, are insufficient for predicting drug sensitivity.
• Deep learning models show promise but face challenges in clinical translation.

Purpose of the Study:

• To develop an advanced computational strategy, DD-Response, for accurate cell-line drug response prediction.
• To overcome limitations in existing models by integrating multiple datasets and improving feature representation.
• To enhance the exploration of mechanisms underlying drug response and facilitate clinical applications.

Main Methods:

• Integrated multiple cell-line drug response datasets using source-specific label binarization to broaden the model's training domain.
• Developed a novel two-dimensional structurized gridding map (SGM) for cell lines and drugs to prevent feature correlation neglect and information loss.
• Constructed a dual-branch, multi-channel convolutional neural network (CNN) for pairwise response prediction.

Main Results:

• DD-Response achieved superior performance in predicting cell-line drug response compared to existing methods.
• The model effectively captured characteristic variations among cell lines and identified key factors influencing drug sensitivity.
• DD-Response demonstrated scalability and potential for predicting clinical patient responses to drug therapies.

Conclusions:

• DD-Response offers a powerful tool for predicting drug response and elucidating underlying molecular mechanisms.
• The strategy is expected to significantly advance drug discovery, repurposing, resistance reversal, and therapeutic optimization.
• This approach holds promise for improving personalized medicine by bridging the gap between cell-line and clinical drug response prediction.