Biological Context-Informed and Population-Stratified Strategies Improve Genetic Diagnosis of CCDC22-Related Disorder

The CCDC22 gene is a key regulator of endosomal trafficking and NF-κB signaling, and its dysfunction is implicated in a spectrum of X-linked neurodevelopmental disorders, including Ritscher-Schinzel syndrome 2, intellectual disability, and epilepsy. Despite the identification of CCDC22 variants having increased, a comprehensive biological characterization, including its spatiotemporal expression and functional molecular networks, remains to be systematically delineated. Furthermore, the rarity of CCDC22 variants created a significant challenge in distinguishing their pathogenicity. This study aimed to explore the integration of biological context-informed and clinical genetic analysis workflows and to perform an exploratory evaluation of prediction algorithms to better optimize the genetic diagnosis process for CCDC22.

Multilevel biological analysis was performed to evaluate the spatiotemporal expression patterns of CCDC22 across developmental stages. A protein-protein interaction network was constructed to identify key functional modules and pathway enrichments. Additionally, an expert-classified dataset of CCDC22 missense variants was utilized to conduct an exploratory performance evaluation of twenty prediction algorithms, specifically assessing the effect of filtering out variants observed as population hemizygotes.

CCDC22 exhibits distinct spatiotemporal expression dynamics and chromatin accessibility patterns closely associated with neurodevelopmental processes. PPI and functional enrichment analyses highlighted its core involvement in endocytic recycling and vesicle transport. In the exploratory missense variant evaluation, meta-predictors, notably ClinPred and MetaRNN, demonstrated the highest predictive potential within this limited cohort. After filtering out variants observed as hemizygotes in the gnomAD database, most of the algorithms' performance improved in distinguishing the pathogenicity of variants and genetic diagnosis. ClinPred, M-CAP, MetaRNN, and SIFT achieved the highest balanced accuracy (84.6%, 80.8%, 78.7%, and 76.1%, respectively). ClinPred, M-CAP, and MetaRNN achieved the highest AUC value (> 0.9).

This study delineates the spatiotemporal and functional molecular network of CCDC22 in neurodevelopment. A combination of population-based strategy and prediction enhanced the performance of most algorithms. ClinPred and MetaRNN showed higher predictive potential. This study may provide insights into the evaluation of variants in CCDC22-related diseases.