An outlier approach: advancing diagnosis of neurological diseases through integrating proteomics into multi-omics guided exome reanalysis

Abstract

Neurodevelopmental disorders (NDDs) often have unknown genetic causes. Current efforts in identifying disease-related genetic variants using exome or genome sequencing still lead to an excessive number of variants of uncertain significance (VUS). There is an increasing interest in transcriptomics and, more recently, proteomics for variant detection and interpretation. In this study, we integrated quantitative liquid chromatography-mass spectrometry proteomics, RNA sequencing, and exome reanalysis to resolve VUS and detect novel causal variants in 34 patients with undiagnosed NDDs, using the software PROTRIDER and DROP to detect protein outliers and RNA outliers, respectively. We obtained a diagnosis in 11 cases (32%) resulting from the increased amount of information provided by the two additional levels of omics (n = 5) and the updated literature evidence (n = 6). Our experience suggests the potential of this outlier-detection multi-omics workflow for improving diagnostic yield in NDDs and other rare disorders.

Fig. 1. Overview of the study.

A Diagnostic workflow of our multi-omics pipeline incorporating exome reanalysis, transcriptomics, and proteomics. B Diagnostic yield of this study. The pie chart on the left shows the proportion of recruited cases that were resolved in this study. The pie chart on the right shows the number of resolved cases whose resolution was guided by different types of omics analysis. RNA-seq and proteomics: cases where the evidence was found in both RNA-seq and proteomics data. Proteomics-only: Cases where the evidence was found only in proteomics data. RNA-seq-only: Cases where evidence was found only in RNA-seq data. Updated literature: Cases resolved through exome reanalysis using evidence from up-to-date literature, in which 5 cases involved missense variants that may alter protein functions but not affect RNA or protein levels significantly, and hence not assessed by our multi-omics approach for detecting expression outliers.

Fig. 2. Percentage of genes from different disease gene panels that were detected by RNA-seq and proteomics.

The disease gene panels were retrieved from publicly available databases and from the literature. n = number of genes in the disease gene panel. Error bars indicate the 95% confidence intervals for each percentage of genes detected.

Fig. 3. Identification of the splice region MSTO1 variant in participant SF185.

The MAE pipeline detected that one of the alleles with a missense variant c.971C>T in MSTO1 has significant monoallelic expression (82%), suggesting a variant triggering nonsense-mediated decay in the other allele. A heterozygous splice region variant c.967-3C>A was identified in the other allele, leading to exon elongation with a putative starting point at c.966+65 and a premature termination codon at c.966+104. Reinspection of the expression and splicing results led to the finding that the exon elongation is aberrant (nominal p-value of 0.0258 and an effect size ΔJ of −0.2) and causes a global reduction in expression (OUTRIDER: fold change = 0.71, Z-score = −4.13, nominal p-value = 1.13 × 10⁻⁴). This case highlights the importance of RNA-seq results reevaluation guided by genetic or clinical data, as for both expression and splicing analyses, the FDR was above the significance cutoff of 0.1.

Fig. 4. SHMT2 protein-only outlier in participant SF269.

Detection of SHMT2 protein-only outlier guided the identification of two compound heterozygous SHMT2 missense variants, which lead to protein destabilization but not transcriptional dysregulation.

Fig. 5. GARS1 protein-only outlier in participant SF231.

Multi-omics analysis suggests the reprioritization of 4 genes, including the protein-only outlier GARS1, in the exome reanalysis of the case. This gene was not detected as a significant RNA expression outlier (FDR > 0.1), yet with a fold change of 0.81 and a nominal p-value of 0.052. A likely pathogenic c.258_259insGTGGCTGAGCTCAAAGC; p.(Pro87ValfsTer9) insertion variant leading to a frameshift was then identified, which explains the individual’s phenotype. This variant was missed in the initial exome analysis.

Fig. 6. Identification of biallelic RNU7-1 variants in participant SF188.

Overexpression of 26 histone genes guided the identification of biallelic RNU7-1 variants reported in individuals with Aicardi–Goutières syndrome, which resulted in misprocessing of the canonical histone transcripts, leading to phenotypes that align with those observed in our patient.

Fig. 7. Identification of 104.5 kb deletion in participant SF197.

Detection of expression outliers of MSFD1 and GFM1 guided the identification of a 104.5 kb deletion covering 11 candidate enhancers associated with GFM1, explaining the individual’s phenotypes.