Rosace: a robust deep mutational scanning analysis framework employing position and mean-variance shrinkage

Area of Science:

• Computational biology
• Statistical genetics
• Protein engineering

Background:

• Deep mutational scanning (DMS) is a powerful technique for assessing the functional impact of thousands of genetic variants in proteins.
• Classical statistical methods struggle with the small sample sizes inherent in DMS experiments, leading to inaccurate p-value calibration when variants are treated independently.

Purpose of the Study:

• To develop a robust statistical framework for analyzing growth-based DMS data that addresses the limitations of existing methods.
• To improve statistical power and control the false discovery rate in DMS analyses.

Main Methods:

• Proposed Rosace, a Bayesian framework that incorporates amino acid position information.
• Implemented parameter shrinkage to share information across variants, enhancing statistical power.
• Developed Rosette for simulating DMS data to assess distributional properties.

Main Results:

• Rosace demonstrates increased power compared to existing tools for analyzing DMS data.
• The framework effectively controls the false discovery rate by leveraging information across parameters.
• Rosace shows robustness even when underlying model assumptions are violated.

Conclusions:

• Rosace offers a more powerful and statistically sound approach for analyzing growth-based DMS data.
• The Bayesian framework provides improved accuracy in identifying functionally significant protein variants.
• Rosace represents a significant advancement in the statistical analysis of high-throughput mutagenesis experiments.