PMC

On each curve, "X" indicates the number of DS genes detected for the FDR of 0.05. Model full - comparison of 4 control samples versus 3 knock-down. Model full paired - comparison of 2 versus 2 paired-end samples. Model full 2 - as model full but including the information about library layout (no results for DRIMSeq because currently, it is not able to fit models with multiple covariates).

DRIMSeq was run with different dispersion estimation strategies: common dispersion and genewise dispersion with no moderation (genewise_grid_none), moderation to common dispersion (genewise_grid_common) and moderation to trended dispersion (genewise_grid_trended). Results presented for Drosophila melanogaster and Homo sapiens simulations with DTU (nonull) and no differential gene expression (node) using transcript counts from kallisto and exonic counts from HTSeq. Additionally, filtered counts (kallistofiltered5, htseqprefiltered5) are used. When the achieved FDR is smaller than the threshold, circles are filled with the corresponding color, otherwise, they are white.

A: Concordance between sQTLseekeR and DRIMSeq. "X" indicates number of tuQTLs for FDR = 0.05. Panel B, C and D show characteristics of tuQTLs and genes detected by sQTLseekeR or DRIMSeq for FDR = 0.05. Values in brackets indicate numbers of tuQTLs or genes in a given set. Dark gray line corresponds to tuQTLs or genes that were identified by both of the methods (overlap). B: Distance to the closest exon of intronic tuQTLs. The light gray line (non_sqtl) corresponds to intronic tuQTLs that were not called by any of the methods. C: Distribution of mean gene expression for genes that are associated with tuQTLs. D: Distribution of the number of expressed transcripts for genes that are associated with tuQTLs. The light gray lines (all_genes) represent corresponding features for all the analyzed genes.

In the first scenario, all genes have the same (common) dispersion, and in the second one, each gene has a different (genewise) dispersion. All genes have expression equal to 1000 and 3 or 10 features with the same proportions estimated from kallisto counts from Kim et al. data set. For each of the scenarios, common, genewise, with and without moderation to common dispersion is estimated with maximum likelihood using the dirmult R package, the raw profile likelihood and the Cox-Reid APL. A: Absolute error of concentration γ ₊ estimates. B: False positive (FP) rate for the p-value threshold of 0.05 of the null two-group comparisons based on the likelihood ratio statistics. Dashed line indicates the 0.05 level. C: Median raw error of γ ₊ estimates. D: Distributions of p-values of the null two-group comparisons based on the likelihood ratio statistics. Additionally, results when true concentration estimates are used are indicated with the gray color.