Simulated RNA-seq data

Simulated RNA-seq data shows that histograms from p value sets with around one hundred  true effects out of 20,000 features can be classified as 'uniform'. RNA-seq data was simulated with polyester R package (Frazee, 2015) on 20,000 transcripts from human transcriptome using grid of 3, 6, and 10 replicates and 100, 200, 400, and 800 effects for two groups. Fold changes were set to 0.5 and 2. Differential expression was assessed using DESeq2 R package (Love, 2014) using default settings and group 1 versus group 2 contrast. Effects denotes in facet labels the number of true effects and N denotes number of replicates. Red line denotes QC threshold used for dividing p histograms into discrete classes. Workflow and code used to run this simulation is available on rstats-tartu/simulate-rnaseq.

Files

de_simulation_results.csv -- merged and processed DE analysis results of simulated data.

simulate-reads-2021-01-25.tar.gz -- raw DE analysis results on 20,000 transcripts from human transcriptome using grid of 3, 6, and 10 replicates and 100, 200, 400, and 800 effects for two groups. Fold changes were set to 0.5, 1, and 2. Differential expression was assessed using DESeq2 with default settings.

simulate-rnaseq.tar.gz -- snakemake workflow and input fasta file to simulate RNA-seq data with polyester and analyse results with DESeq2. Adjust settings in config.yaml to customise simulation. Includes software to run workflow on Linux, given that Conda and snakemake are installed.

The simulate-rnaseq.tar.gz archive can be re-executed on a vanilla machine that only has Conda and Snakemake installed via:

tar -xf simulate-rnaseq.tar.gz
snakemake --use-conda -n