662 - Developing epigenetic models to address the challenge of variants of uncertain clinical significance for pediatric rare diseases

Publication Number: 4648.662

Jian Wang, Shanghai Jiao Tong University School of Medicine, Xuhui, Shanghai, China (People's Republic); Zhengchang Li, Shanghai Jiao Tong University, Shanghai, Shanghai, China (People's Republic); Niu Li, Shanghai Jiaotong Unoversity School of Medicine, Xuhui, Shanghai, China (People's Republic)

Background: Variants in coding genes related to epigenetic regulation can lead to disease-specific methylation profiles across the entire genome, and have been proven to be useful for molecular diagnosis of various neurodevelopmental disorders. However, such diagnostic models and their application value have not yet been constructed and validated in rare disease patients in China.

Objective: We aimed to identify disease-specific epi-signatures in Chinese patients, develop diagnostic models, and conduct clinical diagnostic efficacy evaluations.

Design/Methods: We collected 201 cases from 6 distinct disorders, including Sotos, Kabuki, Rett, CHARGE, Williams-Baure, and Prader-Willi syndrome (PWS), with no less than 30 patients for each group. Meanwhile, 40 age- and gender- matched healthy individuals were collected as controls. 3/4 of the patients were used as the "training cohort", and the whole genome methylation profile was analyzed using the Illumina Infinium Methylation EPIC v2.0 kit. After multiple filtering, disease-specific epi-signatures were constructed, and a classification diagnostic model was established based on support vector machine (SVM) training classifier. The remaining cases were used as the "test cohort" to evaluate the clinical diagnostic efficacy of the models.

Results: A total of 634 probes (100-125/disease) were screened to construct the classification diagnostic model for each disease. 85% of the probes are disease specific, with only 4 probes present in each disease. Among the 56 cases of the test cohort, 51 (91%) obtained diagnostic conclusions that met expectations, including 44 cases with consistent clinical and genotype diagnoses. 7 variants of uncertain significance (VUS) were re-classified, including 1 likely pathogenic variant (c.413+2dup of MECP2) and 6 likely benign variants (p.R1922C of CHD7, p.D1111H, p.G1245A and p.K144Rfs*21 of NSD1, and p.R5340* and c.4020+2T>A of KMT2D). We also tested KDM6A related Kabuki samples and found that the KMT2D-Kabuki model is not suitable for the recognition of KDM6A variants. Further analysis revealed that PWS is more suitable to be modeled separately from other diseases.

Conclusion(s): This study explores the efficacy of whole genome methylation analysis in the diagnosis of Chinese genetic disease patients. The models we constructed show the potential of differential diagnosis. These localized diagnostic models not only provide new molecular basis for VUS interpretation, but also has guiding value for cases with highly matched genotypes but significantly inconsistent clinical symptoms.

Fig 1. Heatmap to display differential methylation profiles between each disease group and control group


Fig.2. Testing samples to verify the diagnostic performance of each model.
(A-D) Sotos, KTM2D-Kabuki, WBS, and Rett models can effectively distinguish between positive (P or LP variants) and negative (B or LB variants) samples. In addition, satisfactory differential diagnostic performance was also demonstrated among the four models. (E) At least two CHAGRE patients with a clear diagnosis were mistakenly diagnosed as negative, and one sample was mistakenly identified as Rett syndrome. (F) Kabuki syndrome with different genotypes exhibit distinct epi-signature. 18 patients with KDM6A mutation using the KMT2D Kabuki model, of which only 4 were determined to be positive.

Fig3. PWS is modeled separately from other diseases.
(A) When modeled together with other diseases, the diagnostic performance of the PWS model is highly unsatisfactory. (B) When the PWS disease is modeled independently (with probe selection not considering differences from other diseases), satisfactory diagnostic performance is achieved.