Deficiency of adenosine deaminase (ADA or ADA1) has broad clinical and genetic heterogeneity. Screening techniques can identify asymptomatic infants whose phenotype and prognosis are indeterminate, and who may carry ADA variants of unknown significance. We systematically assessed the pathogenic potential of rare ADA missense variants to better define the relationship of genotype to red blood cell (RBC) total deoxyadenosine nucleotide (dAXP) content and to phenotype. We expressed 46 ADA missense variants in the ADA-deficient SØ3834 strain of Escherichia coli and defined genotype categories (GCs) ranked I to IV by increasing expressed ADA activity. We assessed relationships among GC rank, RBC dAXP, and phenotype in 58 reference patients with 50 different genotypes. We used our GC ranking system to benchmark AlphaMissense for predicting variant pathogenicity, and we used a minigene assay to identify exonic splicing variants in ADA exon 9. The 46 missense variants expressed ∼0.001% to ∼70% of wild-type ADA activity (40% had <0.05% of wild-type ADA activity and 50% expressed >1%). RBC dAXP ranged from undetectable to >75% of total adenine nucleotides and correlated well with phenotype. Both RBC dAXP and clinical severity were inversely related to total ADA activity expressed by both inherited variants. Our GC scoring system performed better than AlphaMissense in assessing variant pathogenicity, particularly for less deleterious variants. For ADA deficiency, pathogenicity is a continuum and conditional, depending on the total ADA activity contributed by both inherited variants as indicated by GC rank. However, in patients with indeterminate phenotype identified by screening, RBC dAXP measured at diagnosis may have greater prognostic value than GC rank.