Simulated RNA-seq data for differential splicing analysis with covariates

The repository includes alignments of simulated RNA-seq data for evaluating differential splicing detection with covariates. Starting from an empirical transcript expression matrix trained on an RNA-seq data set from lung fibroblasts (GenBank A# SRR493366) and using GENCODE v.41 as reference, 11.5 million 100 bp long paired-end reads were generated per sample, from 2,000 genes with two or more expressed isoforms. RNA-seq data was simulated for one ‘condition’, with values ‘control’, ‘disease’ and ‘stage2’, with one covariate, ‘biological sex’, with values ‘M’ and ‘F’.  5 samples each were simulated for each (condition x sex) category. Changes were simulated in the expression (DE) and/or the splicing ratio (DS) of genes as follows. Changes in expression (DE) were simulated by either halving or doubling the expression level of the gene. Changes in splicing ratios (DS) were simulated by swapping the expression levels of the gene’s top two transcript isoforms. All RNA-seq data was mapped to the hg38 genome with the spliced alignment tool STAR v2.7.10a.

 Pairwise comparison alignment set: Differences due to ‘condition’ between two states, ‘control’ and ‘disease’, were simulated at 600 genes, including 200 DE, 200 DS and 200 DE+DS genes. Differences in ‘biological sex’ (covariate) were represented as changes in 300 genes, including 100 from each of the DS, DE and DE+DS categories. Hence, the target gene set for differential splicing ratio (DSR) pairwise comparisons consists of the pooled 200 DS and 200 DS+DE genes differentially spliced between the ‘control’ and ‘disease’ states, while for differential splicing abundance (DSA) pairwise comparisons the target gene set is the set of 600 modified genes, 200 in each of the DS, DE and DS+DE categories.

 Multiway (3-way) comparison alignment set: Additional changes between ‘disease’ and ‘stage2’ were made to 100 of the previously modified genes, as well as to a set of 200 additional genes not encountered previously, for each of the categories DE, DS and DE+DS. Therefore, for DSR three-way comparisons, the target gene set represents the 800 genes simulated as being DS or DE+DS between any of the ‘control’, ‘disease’ and ’stage2’ categories, while for the multi-way DSA comparisons the target is the full set of 1,200 genes (400 DE, 400 DS and 400 DE+DS) simulated to have changed between any of the 'control', 'disease’ and ‘stage2’ states.

 Further details: See the ‘key’ directories in each package for the gene lists.