Harmonization and integration of pharmacogenomics screens

Area of Science:

• Pharmacogenomics
• Computational Biology
• Drug Discovery

Background:

• Large-scale pharmacogenomic screens provide extensive data on tumor cell line drug responses.
• Challenges exist in integrating and comparing data due to varied experimental platforms and response metrics.
• Existing repositories offer data access but lack standardized analysis protocols.

Purpose of the Study:

• To evaluate different data analysis protocols for harmonizing pharmacogenomic screening data.
• To identify the most effective method for improving data consistency across diverse experimental setups.
• To facilitate more robust large-scale analyses of drug response in cancer cell lines.

Main Methods:

• Surveyed various data analysis protocols, including different curve-fitting functions (sigmoid, piecewise linear).
• Compared different response metrics such as IC50, EC50, and integrated AUC.
• Assessed the impact of different drug concentration windows (full range vs. truncated).

Main Results:

• An area under the curve (AUC) derived from a sigmoidal curve fitted to a truncated dose range demonstrated the highest agreement between screening platforms.
• This harmonization protocol significantly outperformed other methods evaluated.
• The chosen method also improved the alignment of drug responses across repeated experiments on the same platform.

Conclusions:

• A standardized approach using sigmoidal fitting on truncated dose ranges is recommended for harmonizing pharmacogenomic data.
• This method enhances data comparability and reliability for large-scale drug response analyses.
• Improved data integration supports more effective drug discovery and personalized medicine efforts.