Systematic benchmarking of methods to analyze RNA editing in single-cell transcriptomes

Adenosine-to-inosine (A-to-I) RNA editing is widespread in metazoans and functionally important in development and disease, yet a robust framework for studying it in single-cell transcriptomes remains absent. Here, we systematically evaluate strategies to profile A-to-I editing in single-cell RNA sequencing (scRNA-seq) datasets across diverse species and tissue types, totalling >1.1 million cells from 14 publications. Our analyses reveal substantial differences in performance between preprocessing steps, variant callers, and technology platforms. Despite read sparsity in single cells, reliable site discovery is achievable at the cell-type level, with BCFtools plus annotation-based strand assignment emerging as the top overall pipeline. Moreover, single-cell insights are obtainable through Alu Editing Index (AEI) measurements of global editing activity. We further introduce an approach for differential editing analysis that optimally balances between precision and recall. Applying our framework to disease contexts, we find that RNA editing programs are anchored in cell identity but can be reshaped by the tumor microenvironment. In Down syndrome, ADAR2 dysregulation is restricted to a subset of excitatory neurons and associated with age and inflammation. Our best practices code repository is freely available to empower future studies on single-cell RNA editing.