Clinical epigenomics: genome-wide DNA methylation analysis for the diagnosis of Mendelian disorders

Abstract

Purpose: We describe the clinical implementation of genome-wide DNA methylation analysis in rare disorders across the EpiSign diagnostic laboratory network and the assessment of results and clinical impact in the first subjects tested.

Methods: We outline the logistics and data flow between an integrated network of clinical diagnostics laboratories in Europe, the United States, and Canada. We describe the clinical validation of EpiSign using 211 specimens and assess the test performance and diagnostic yield in the first 207 subjects tested involving two patient subgroups: the targeted cohort (subjects with previous ambiguous/inconclusive genetic findings including genetic variants of unknown clinical significance) and the screening cohort (subjects with clinical findings consistent with hereditary neurodevelopmental syndromes and no previous conclusive genetic findings).

Results: Among the 207 subjects tested, 57 (27.6%) were positive for a diagnostic episignature including 48/136 (35.3%) in the targeted cohort and 8/71 (11.3%) in the screening cohort, with 4/207 (1.9%) remaining inconclusive after EpiSign analysis.

Conclusion: This study describes the implementation of diagnostic clinical genomic DNA methylation testing in patients with rare disorders. It provides strong evidence of clinical utility of EpiSign analysis, including the ability to provide conclusive findings in the majority of subjects tested.

Fig. 1. A multiclass supervised classification system.

This classification system, referred to as a methylation variant pathogenicity (MVP) score, can classify samples using the 43 episignatures in EpiSign v2. Shown here are examples using the Cornelia de Lange syndrome (CdLS) signature (top) and Sotos signature (bottom), applied to over 1,500 samples from subjects with various neurodevelopmental syndromes or from healthy controls. In each case, samples from the respective syndrome all have high scores while samples from other syndromes and controls all have low scores, demonstrating the sensitivity and specificity of the classifier. The likely pathogenic variants described in Fig. 2 (CdLS) and 3 (Sotos) are shown here larger and in red.

Fig. 2. DNA methylation (EpiSign) analysis of peripheral blood from two subjects with variants of unknown clinical significance (VUS) in SMC1A, the causative gene for Cornelia de Lange syndrome (CdLS) type 2.

The variants are SMC1A:c.598A>C, p.(Lys200Gln) (labeled red and clustering with CdLS samples) and SMC1A:c.1280A>G, p.(Glu427Gly) (labeled red and clustering with control samples). (a) Hierarchical clustering and (b) multidimensional scaling plot of subjects with a confirmed CdLS episignature, controls, and the VUS under investigation. The SMC1A:c.598A>C VUS clustering with the CdLS samples indicates it has a DNA methylation signature similar to that seen in other CdLS samples and suggesting that the variant is pathogenic, while SMC1A:c.1280A>G is likely benign. (c) Methylation variant pathogenicity (MVP) score, a multiclass supervised classification system capable of discerning between the 43 different episignatures in EpiSign V2, was applied to the SMC1A:c.598A>C likely pathogenic variant (top) and the SMC1A:c.1280A>G likely benign variant (bottom). This classification system generates a probability score for each episignature, with a score near 1 indicating that the sample has an episignature similar to the reference episignature.

Fig. 3. DNA methylation (EpiSign) analysis of peripheral blood from two subjects with variants of unknown clinical significance (VUS) in NSD1, the causative gene for Sotos syndrome type 1.

The variants are NSD1:c.4982G>C,p.Cys1661Ser) (labeled red and clustering with NSD1 samples) and NSD1:c.3331G>T,p.Asp1111Tyr (labeled red and clustering with control samples). (a) Hierarchical clustering and (b) multidimensional scaling plot of subjects with a confirmed Sotos episignature, controls, and the VUS under investigation. The NSD1:c.4982G>C VUS clustering with the Sotos samples indicates it has a DNA methylation signature similar to that seen in other Sotos samples and suggesting that the variant is pathogenic, while NSD1:c.3331G>T is likely benign. (c) Methylation variant pathogenicity (MVP) score, a multiclass supervised classification system capable of discerning between the 43 different episignatures in EpiSign v2, was applied to the NSD1:c.4982G>C likely pathogenic variant (top) and the NSD1:c.3331G>T likely benign variant (bottom). This classification system generates a probability score for each episignature, with a score near 1 indicating that the sample has an episignature similar to the reference episignature.

Fig. 4. Clinical case EpiSign assignment.

Assessment of the notable clinical cases. (a) Principal component analysis (PCA) plots for Clin136 (no genetic variant identified), Clin6 (no genetic variant identified), and Clin187 (DNMT3A:c.2146G>A [p.Val716Ile]). (b) Corresponding methylation variant pathogenicity (MVP) scores for the notable clinical cases. Both PCA plots and MVP scores provide conclusive evidence for EpiSign classification of these clinical cases.