Empirical null estimation using zero-inflated discrete mixture distributions and its application to protein domain...

Area of Science:

• Genomics and Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• Gene-centric mutation analyses have limitations in capturing functional context.
• Protein domain position analysis offers a more functionally relevant approach.
• Simultaneous inference in large-scale mutation studies presents statistical challenges.

Purpose of the Study:

• To develop a statistical procedure for selecting significant mutation counts.
• To control Type I errors using False Discovery Rate (FDR) procedures.
• To enhance statistical power in large-scale mutation analysis.

Main Methods:

• Utilized zero-inflated models, specifically the Zero-inflated Generalized Poisson (ZIGP) distribution.
• Developed data-dependent methods to determine a cut-off value for the null distribution.
• Implemented a two-stage testing procedure involving a screening process.

Main Results:

• The proposed two-stage procedure effectively controls the False Discovery Rate (FDR).
• Methods for determining the cut-off value were presented and illustrated.
• The procedure demonstrated superior empirical power compared to standard methods in simulations and protein domain data analysis.

Conclusions:

• The two-stage testing procedure offers a powerful and statistically sound approach for mutation count analysis.
• This method improves upon existing techniques by better accounting for functional mutation context.
• The approach is effective in identifying significant mutations while maintaining stringent error control.