This research article evaluates the performance of AlphaMissense (AM), a new technology that predicts the pathogenicity of missense variants in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, against cystic fibrosis (CF) variants. AM predicts pathogenicity based on learned protein structure and evolutionary features. The study found that AM generally predicted high pathogenicity for CFTR residues, leading to a high false positive rate and fair classification performance on CF variants from the CFTR2.org database. AM's pathogenicity scores correlated modestly with clinical outcomes such as sweat chloride levels, pancreatic insufficiency rates, and Pseudomonas aeruginosa infection rates, but not with CFTR trafficking and folding competency in vitro. However, AM scores correlated well with CFTR channel function in vitro, demonstrating that the dual structure and evolutionary training approach learns important functional information. The study also found that AM had limited utility in predicting the pharmacological response of CFTR variants, known as theratyping. Overall, AM may be useful for determining if polymorphisms in CFTR are recessive CF variants but requires further development to refine its targeting of emerging precision CF therapeutics.This research article evaluates the performance of AlphaMissense (AM), a new technology that predicts the pathogenicity of missense variants in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, against cystic fibrosis (CF) variants. AM predicts pathogenicity based on learned protein structure and evolutionary features. The study found that AM generally predicted high pathogenicity for CFTR residues, leading to a high false positive rate and fair classification performance on CF variants from the CFTR2.org database. AM's pathogenicity scores correlated modestly with clinical outcomes such as sweat chloride levels, pancreatic insufficiency rates, and Pseudomonas aeruginosa infection rates, but not with CFTR trafficking and folding competency in vitro. However, AM scores correlated well with CFTR channel function in vitro, demonstrating that the dual structure and evolutionary training approach learns important functional information. The study also found that AM had limited utility in predicting the pharmacological response of CFTR variants, known as theratyping. Overall, AM may be useful for determining if polymorphisms in CFTR are recessive CF variants but requires further development to refine its targeting of emerging precision CF therapeutics.