Multilevel models improve precision and speed of IC50 estimates

Area of Science:

• Pharmacology
• Computational Biology
• Biostatistics

Background:

• Drug sensitivity assays on cell lines often yield inaccurate IC50 estimates due to experimental variability.
• Current methods typically analyze drug-cell line responses individually, limiting precision.

Purpose of the Study:

• To improve the accuracy and precision of half-limiting dose (IC50) estimations.
• To develop a novel statistical approach leveraging all available dose-response data simultaneously.

Main Methods:

• A multilevel mixed-effects model was developed to integrate dose-response data across multiple drugs and cell lines.
• This approach contrasts with traditional single drug-cell line analysis.

Main Results:

• The multilevel model demonstrated high concordance with existing Bayesian methods for well-behaved data.
• It significantly outperformed the Bayesian model with noisy or variable data.
• The model reduced extreme IC50 estimates and increased overall precision.

Conclusions:

• The proposed multilevel model offers a superior approach for IC50 estimation compared to conventional methods.
• It provides more precise drug sensitivity measurements and is computationally efficient.
• This method enhances the reliability of drug efficacy assessments in preclinical research.