Leveraging in-silico deep learning and computational analyses to predict the pathogenicity of ROBO4 variants of uncertain significance in aortic aneurysm and dissection patients.

Variants of uncertain significance (VUS) in genes implicated in thoracic aortic aneurysm (TAA) present clinical challenges due to ambiguous pathogenicity and low patient representation. This study investigates the pathogenic potential of missense VUS in the ROBO4 gene, previously associated with vascular integrity and ascending aortic aneurysm. Clinical and genetic data from five patients with heterozygous ROBO4 VUS and thoracic aortic aneurysms or dissections were analyzed. Computational tools including AlphaFold2, AlphaMissense, REVEL, PolyPhen-2, SIFT, FATHMM, MutationTaster2, GranthamMatrix, and PhastCons were utilized to predict pathogenicity and structural impacts. Patients exhibited varying severities of aortic pathology, from elective aneurysm repairs to extensive familial aneurysmal histories. Structural modeling revealed significant differences in residue positions and biochemical properties, particularly for extracellular domain variants affecting critical beta-sheet structures involved in vascular stability. Notably, patient-specific predictions aligned computational evidence with clinical severity, suggesting potential genotype-phenotype correlations. For example, a variant (Q44P) showed strong pathogenic predictions coinciding with severe familial presentations. These computational predictions, validated by clinical data, highlight a novel and efficient workflow for evaluating VUS pathogenicity, informing precision medicine, and guiding counseling for aortic degenerative diseases. Ultimately, we demonstrate the value of integrating computational modeling with clinical data to decipher the pathogenic significance of genetic variants in cardiovascular diseases.