Identification of RNA editing sites and quantification of RNA editing levels are mainly executed by REDItools 2.0 REDItoolDenovo.py (https://github.com/BioinfoUNIBA/REDItools2). Firstly, all candidate RNA editing sites are identified and quantified based on read coverage and variation frequency at each site using Parallel Strategy of REDItools 2.0. Secondly, candidate RNA editing sites located in Alu and non-Alu regions are filtered by different parameter configurations, and annotated subsequently based on a variety of annotation files such as gene annotation, RepeatMasker, SNP and known RNA editing information. Thirdly, additional filtering criteria is used to obtain more accurate novel editing sites located in non-Alu regions because non-Alu region usually has a narrow range of editing sites. To do this, Pblat (Wang and Kong 2019) is used to detect the mismatched and multi-mapping reads, while Samtools (Li et al. 2009) is used to delete duplicated reads. Finally, RNA editing sites are tagged as novel or known sites. In the current version, RNA editing types of both A-to-I and C-to-U are included.
Source of annotation files: (1) Gene annotation file: GENCODE V33, (2) RepeatMasker annotation file: UCSC, (3) SNP annotation files: UCSC, (4) Known RNA editing sites: REDIportal database (). Genomic coordinates of the RepeatMasker file and the known RNA editing sites file are converted from hg19 to hg38 using UCSC liftover.

Software used for read alignments depends on the type of data that will be analyzed. For drop-based single cell RNA-seq data, there is no quality control step in the upstream analysis. For protocols such as Smart-Seq, Smart-Seq2 and Fluidigm C1, processing pipeline is the same as bulk RNA-seq analysis, including fastp, RseQC and RSEM, except a special parameter ‘--single-cell-prior’ using Dirichlet (0.1) as the prior to calculate posterior mean estimates and credibility intervals in the RSEM step. ‘CellRanger’ is implemented for data generated from the 10X Genomics platform. The matrix counting step is slightly more complicated for drop-based single-cell data because the tools need to keep track of where each read came from (which cell and which transcript, if UMI were used). To obtain a matrix of read counts per gene is part of the alignment step, where rows usually correspond to genes and columns to cells.

Manual curation of metadata of all included RNA-seq datasets are done on 2 levels (‘Project’ and ‘Sample’) based on the structured curation model listed below.