Rosace

Rosace employs a Bayesian framework to analyze growth-based deep mutational scanning (DMS) data, leveraging amino acid position information to increase statistical power and control the false discovery rate.

Key Features:

• Position-based shrinkage: Rosace shares information across parameters via shrinkage using amino acid position information to improve power and false discovery rate control.
• Mean-variance shrinkage: Rosace implements shrinkage that shares information across parameters to address small sample sizes that render classical statistical methods ineffective.
• Robustness to model violations: Rosace maintains performance when model assumptions are violated.
• Simulation companion (Rosette): Rosette simulates the distributional properties of DMS data for method evaluation and benchmarking.

Scientific Applications:

• Deep mutational scanning analysis: Rosace analyzes growth-based DMS experiments that measure the effects of thousands of genetic variants in a protein simultaneously.
• Statistical inference for variant effects: Rosace enables calibrated statistical testing for DMS variant effects where classical methods produce poorly calibrated p-values when treating variants independently.

Methodology:

Rosace uses a Bayesian framework with shrinkage estimation that shares information across parameters via amino acid position information and mean-variance relationships. Rosette simulates DMS data with appropriate distributional properties.