Power calculator for detecting allelic imbalance using hierarchical Bayesian model.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Allelic imbalance (AI) refers to the differential expression of alleles in diploid organisms.
• AI can be influenced by various factors including tissues, treatments, and environmental conditions.
• Existing methods for AI detection need robust statistical frameworks for evaluating type I error and power, especially concerning the trade-off between sequencing reads and biological replicates.

Purpose of the Study:

• To develop and evaluate methods for estimating type I error and statistical power in detecting allelic imbalance (AI).
• To determine the optimal balance between sequencing depth (reads) and the number of biological replicates for robust AI detection.
• To provide a computational tool for simulating AI scenarios and assessing statistical power.

Main Methods:

• Simulated allelic-specific read data under various scenarios of AI.
• Calculated type I error and statistical power based on simulated data.
• Developed and utilized a Python package for AI data simulation and power analysis.

Main Results:

• Detecting a 10-30% deviation from allelic balance within a condition requires 240-2400 allele-specific reads distributed across 3-12 replicates for >80% power.
• Detecting a 20-30% difference in AI between conditions requires 240-960 allele-specific reads across 8 replicates.
• Increasing the number of replicates enhances statistical power more effectively than increasing sequencing coverage without impacting type I error.

Conclusions:

• The number of biological replicates is a more critical factor than sequencing depth for achieving adequate statistical power in AI studies.
• The developed Python package facilitates the estimation of type I error and power, aiding in experimental design for AI detection.
• This work provides essential guidance for researchers designing experiments to study allelic imbalance.