Systematic evaluation of the two main blood-based RNA-seq approaches for Mendelian disease diagnosis

Abstract

Background: As an adjunct to diagnostic exome sequencing and whole genome sequencing, RNA sequencing (RNAseq) has been demonstrated to improve diagnostic yield for Mendelian diseases. However, systematic evaluation of the associated experimental and computational processes and the establishment of robust and efficient practices for RNA diagnostics implemented in the clinic to analyse readily accessible whole blood samples are still required. Methods: We simulated clinical conditions in which each patients sample is tested only once, and we evaluated the two typical experimental protocols (polyA selection and rRNA depletion) by comparing the expression profiles, aberrant splicing events and monoallelic expression (MAE) identified from 11 patients in clinical settings with different bioinformatics software. Results: We demonstrated that a higher proportion of unique reads from polyA selection than rRNA depletion were mapped to exons or exon_intron junction regions (84.54% vs. 40.14%), resulting in more detectable OMIM genes (TPM > 1) in the blood (65.29% vs. 59.79%); thus, the rRNA depletion method requires a median of 258 more valid reads per gene to achieve the same level of gene quantification. Moreover, although the transcriptome profiling of protein-coding genes in the two methods is highly correlated, polyA-selection offers more sensitive detection of MAE variants and aberrant splicing under common filtering conditions in combination with DROP. Conclusions: A combination of polyA+ and DROP is recommended when implementing blood-based RNA-seq for the diagnosis of Mendelian diseases in clinical practice, and filtering criteria for aberrant expression, aberrant splicing and MAE variants are suggested for reference.