Cracking the Code: Genotype-Phenotype Correlation Models in Sarcoglycanopathies.

OBJECTIVE

Sarcoglycanopathies are among the most severe limb-girdle muscular dystrophies (LGMD), though milder presentations have been described. These diseases are primarily caused by missense variants, but the limited predictability of their effect on protein maturation, complex formation, and transport has hindered reliable genotype-phenotype correlations. This study aimed to establish accurate genotype-phenotype correlations for LGMDR3, LGMDR4, and LGMDR5.

METHODS

We analyzed the largest sarcoglycanopathy cohort to date (n = 541). Clinical data, including age at symptom onset and loss of ambulation, were collected and used to define phenotype severity. Predictive models were developed considering the impact of non-truncating variants on secondary structure, functional domains, evolutionary conservation, and intra-complex protein-protein interactions. Patients carrying two variants predicted to disrupt membrane trafficking were expected to present with severe phenotypes.

RESULTS

For LGMDR3, the best-performing model classified variants affecting β-sheets, cadherin-like, and ATP-binding domains as disruptive to membrane translocation, achieving 89% predictive power, 0.867 balanced accuracy, and 2.4% false-negative rate (clinically severe patients who were wrongly classified by the model as mild). For LGMDR4 and LGMDR5, the best-performing model involved conserved residues, β-sheets, EGF-like domain, and protein-protein interactions-reaching 80% predictive power, 0.689 balanced accuracy, and 3.1% false-negative rate (LGMDR4), and 90% predictive power, 0.536 balanced accuracy, with no false negatives (LGMDR5). Additionally, we developed an open-access predictive tool for clinical and research application.

INTERPRETATION: This study provides a robust genotype-phenotype correlation for sarcoglycanopathies, improving prognosis, patient management, and clinical trial recruitment.