Pseudo-precision in gene expression values can reduce efficiency.

Area of Science:

• Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• Microarray analysis involves low-level analysis to estimate gene expression from scanned images.
• Algorithms can produce gene expression values with excessive decimal places, leading to pseudo-precision.
• Pseudo-precision results in a lack of tied values, complicating statistical analyses.

Purpose of the Study:

• To investigate the impact of pseudo-precision on statistical tests for gene expression analysis.
• To propose and evaluate a method for handling pseudo-precision by rounding expression values.
• To enhance the efficiency of statistical tests in detecting differentially expressed genes.

Main Methods:

• Suggested omitting pseudo-precision by rounding computed gene expression values.
• Utilized average ranks for nonparametric tests when ties occurred after rounding.
• Applied the Wilcoxon rank sum test to two actual data sets.

Main Results:

• Rounding gene expression values resulted in a more efficient statistical test.
• The average p-value was decreased after rounding.
• The number of p-values smaller than 0.05 increased with rounding.

Conclusions:

• Random noise from pseudo-precision can decrease the efficiency of statistical tests for differential gene expression.
• Rounding gene expression values is a viable method to mitigate the negative effects of pseudo-precision.
• The findings are applicable to various fields dealing with digital data precision.