Recent large-scale global genome screening highlighted genes having extensive amounts of variants of unknown significance (VUS). Among these genes, ABCA4, ADGRV1, RP1L1 and USH2A are linked to highly prevalent and progressive retinal degenerative diseases, notably, retinitis pigmentosa, occult macular dystrophy, Usher Syndrome and Stargardt disease. In order to better characterize these unknown variants, we have used protein sequence and structural information to identify the pathological drivers of clinically-observed missense mutations across these genes. In doing so, we have computed the effect of missense mutations on protein stability and interactions with other proteins and ligands. We have additionally obtained variant effect calculations from evolutionary deep learning models to cover a wide range of potential effects. Using these insights, we have developed a machine learning-based classifier capable of detecting known disease variants across different retinopathies, while also able to prioritize current VUS according to disease risk. Clinically, this has important applications in patient monitoring and treatment.