XAEM

XAEM is a novel computational method for estimating isoform-level gene expression from RNA-seq data. Unlike existing methods that rely on simplifying assumptions and known biases, XAEM employs a more flexible and robust statistical model called the bilinear model. By jointly estimating the design matrix and expression levels from multi-sample RNA-seq data, XAEM automatically corrects for potentially unknown biases.

The method utilizes an alternating expectation-maximization (AEM) algorithm and quasi-mapping for read alignment, resulting in a fast and accurate tool. XAEM demonstrates improved performance compared to recent advanced methods, particularly for multiple-isoform genes in simulated datasets, and achieves better rediscovery rates in independent validation sets for differential expression analysis of real single-cell RNA-seq data.