Statistical methods for comparing two independent exponential-gamma means with application to single cell protein...

Area of Science:

• Genomics
• Bioinformatics
• Statistical Modeling

Background:

• Log transformation is standard for genomic data skewness, often assuming normality.
• Traditional statistical tests (e.g., two-sample t-test) may yield false positives with log-transformed data.
• The choice of distribution significantly impacts genomic data analysis validity.

Purpose of the Study:

• Introduce the Exp-gamma distribution as a model for log-transformed single-cell protein abundance data.
• Highlight the limitations of the two-sample t-test and Wilcoxon-Mann-Whitney (WMW) test for this data type.
• Propose and evaluate new statistical inference methods for Exp-gamma distributed data.

Main Methods:

• Presented the Exp-gamma distribution for log-transformed protein abundance data.
• Demonstrated the inadequacy of two-sample t-test and WMW test.
• Developed and assessed statistical inference techniques for hypothesis testing and confidence intervals.

Main Results:

• The Exp-gamma distribution is proposed as a suitable model for log-transformed protein abundance data.
• Two-sample t-test and WMW test show limitations in analyzing such data.
• Novel statistical methods for Exp-gamma distributed data were evaluated.

Conclusions:

• The Exp-gamma distribution offers a more appropriate statistical framework for analyzing log-transformed protein abundance in single-cell experiments.
• Standard statistical tests are not suitable for this data type.
• The proposed inference methods provide a valid approach for hypothesis testing and confidence interval estimation.