Two alleles of a diploid individual can show significantly different expression referred to as allelic imbalance (AI). Comparing AI between conditions or tissues can provide new insights into the mechanisms of gene expression regulation. There is extensive literature demonstrating Type I error in AI studies is high, particularly when failing to account for map bias and/or using a binomial test. BayesASE_power was developed to address the current lack of understanding of the power for comparing AI between conditions and, in particular, what is the best allocation of resources for boosting power.
BayesASE_power consists of tools that enable users to simulate RNA-seq read counts with a previously published Bayesian model of AI as implemented in BayesASE with any number of biological replicates, reads, and levels of AI. It aggregates the results across multiple simulated datasets to estimate and compare Type I error and power.
- R >= 3.6.1 and "here" package
- python3 with pandas-1.2.4, matplotlib-3.4.1, seaborn-0.11.1, and numpy-1.18.1
A sample script to run the software is provided in the scripts/ folder.