Script explanations below:
MBASED_F1.rmd contains all preprocessing, filtering, and statistics. Use this first to prepare data, and to calculate global reference bias, and to estimate rho for running MBASED using a beta-binomial distribution.
Run_MBASED.R and Run_MBASED_Beta.R are the scripts for actually generating results from MBASED.
The former assumes a binomial distribution of read counts and the latter assumes a beta-binomial distribution.
MBASED_F1_results.rmd contains summary functions, results and analysis
F1_kallisto.sh was used to trim and then pseudo-align the raw reads using trimmomatic and kallisto
subset_kallisto.R contains a modified helper function subset_kallisto() from a defunct pull request for sleuth. This was needed to subset kallisto results, so that we could check for DE using ONLY genes that contain SNVs, for comparison to ASE results.
sleuth uses kallisto bootstraps to check for differential expression while taking inferential variance (from the EM algorithm) into account. Our standard lab DE pipeline uses edgeR so I also included that.
sleuth_F1.rmd contains sleuth creation and quality control. Use sleuth_live(so) to look at the results.
edgeR_F1.rmd was used to check for DE using a second method, again with ONLY genes that contain SNVs
ASE_Visualization is a shiny app for visualizing the estimated kallisto read counts in each sample and the pooled allele specific reads for any given gene. You can choose which list of genes to look at: genes with ASE, DE, both, or all genes.
vcf_prep.rmd Filtering and prepping VCF data for downstream QTL analysis
aseQTL_subset.rmd uses 100 randomly sampled SNVs to do 1000 permutations on, for calculating LOD thresholds.
To do: examine technical variance in DE genes called by edgeR but not sleuth (done but not reported yet)