A statistical framework for analyzing deep mutational scanning data.

Area of Science:

• Molecular Biology
• Genomics
• Biophysics

Background:

• Deep mutational scanning (DMS) is a powerful technique for assessing the functional impact of protein sequence variations.
• Accurate quantification of variant effects is crucial for understanding protein function and evolution.
• Existing DMS methods often lack robust error estimation, potentially leading to the inclusion of noisy data.

Purpose of the Study:

• To develop and validate a novel statistical model for deep mutational scanning that incorporates error estimation.
• To improve the reliability of DMS data analysis by accounting for sampling error and replicate consistency.
• To enhance the ability to identify functional protein variants and perform robust hypothesis testing.

Main Methods:

• Development of a new statistical model for DMS data analysis.
• Integration of error estimation for each variant measurement, considering sampling error and replicate concordance.
• Application and validation of the model on six DMS datasets (one novel, five published) covering 243,732 variants.
• Simulations to assess model performance under various experimental conditions (e.g., cell growth, binding assays, common errors).

Main Results:

• The new model provides reliable error estimates for DMS measurements.
• Demonstrated superiority in identifying and removing noisy variants compared to existing methods.
• Improved performance in hypothesis testing for functional variant identification.
• Successful application across diverse DMS datasets and experimental setups.

Conclusions:

• The developed statistical model significantly enhances the accuracy and reliability of deep mutational scanning data analysis.
• The model's ability to estimate errors and handle experimental noise empowers researchers with more robust insights into protein variant function.
• Implementation in Enrich2 software facilitates broader adoption and application by the research community.