Fig. 3
From: Clinical implementation of RNA sequencing for Mendelian disease diagnostics
Aberrant expression. A Distribution of genes per sample that were detected as expression outliers, for all genes and genes known to cause a disease (OMIM), stratified by outlier class. B Observed over expected number of overexpression and underexpression outliers (y-axis, log-scale) for loss-of-function intolerant genes, OMIM genes, and mitochondrial disease genes (x-axis). Error bars represent 95% confidence intervals of pairwise logistic regressions. C Gene expression fold change relative to the OUTRIDER-modeled expected value of all disease-causal genes that were aberrantly expressed in their corresponding affected sample. Each dot corresponds to a sample, with the affected ones in red. Data stratified by cases diagnosed via RNA-seq (n = 25) and diagnosed via WES (n = 22). Genes with a dominant mode of inheritance are marked with a * (n = 3). The two NDUFA10 cases are siblings, as well as the two DNAJC3 cases. The three TIMMDC1 cases are unrelated. D Gene-level significance (−log10(P), y-axis) versus Z-score, with UFM1 labeled among the expression outliers (red dots) of sample R20754. E Schematic depiction of the NM_016617.2:c.-273_-271del UFM1 deletion (red rectangle) detected by WES in sample R20754. Figure not shown at genomic scale. F Fraction of recalled underexpression outliers simulated with different fold changes (depicted in shades of blue) per mean gene expression (measured in raw read counts). Recall was computed in 50-wide intervals and dots are depicted in the center of the intervals. At a mean read count of 450 (vertical red dashed line), half of the simulated outliers with a fold change of 0.5 are recalled, allowing for investigating dominant genes and compound heterozygotes genes with a single downregulated allele. G Proportion of genes expressed at a given mean expression or higher, colored by different gene classes. Genes are taken from the GENCODE annotation, release 29 (Methods). A total of 9656 genes (16%), 9325 protein coding (46%), and 2098 OMIM genes (55%) have a mean read count higher than 450