AMDBNorm powers batch effects correction for RNA-seq count data

Batch effects are key challenges in RNA sequencing (RNA-seq) data analysis and are often caused by technical differences due to a variety of factors such as experimental conditions, sample processing procedures, sequencing platforms, and so on. Variations caused by these factors obscure biological signals, which can affect the accuracy and comparability of data. Although a number of batch correction algorithms have been developed, they still have certain limitations in dealing with the complexity of RNA-seq data and in avoiding loss of biological signals. In this study, we develop the application of adjustment mean distribution-based normalization (AMDBNorm) algorithm on batch correction of RNA-seq data. AMDBNorm is a probability distribution-based algorithm that eliminates technical differences while preserving biological signals by aligning different batches of data into a single reference batch. We investigate the effectiveness of AMDBNorm in batch effects removal of RNA-seq data with several real datasets. It is found through uniform manifold approximation and projection (UMAP) visualization and a variety of quantitative evaluation tools such as BatchQC, Principal Analysis of Variance (PVCA), and k-nearest-neighbor batch-effect test (kBET) that AMDBNorm can effectively retain biological signals while removing batch effects, and its performance is superior to the existing algorithms in several metrics. Therefore, AMDBNorm provides a new and effective tool for batch correction of RNA-seq data, which helps to improve the data interpretation and the comparability of experimental results in transcriptomic studies.