Abstract Abstract 274 The ABO(H) carbohydrate blood group system is highly polymorphic in humans, with >200 ABO alleles described to date. The ABO gene encodes the glycosyltransferase, ABO, which has two main isoforms which generate blood group A or blood group B. Blood group O is the result of a non-functional ABO glycosyltransferase. ABO blood group is commonly imputed in genome-wide association studies where ABO is associated with phenotypes such as von Willebrand Factor level or myocardial infarction. We sought to determine the accuracy of a common ABO genotyping algorithm in a repository of 1092 individuals from a healthy sibling cohort study, the Genes and Blood Clotting Cohort (GABC). We utilized three published single nucleotide variants (SNVs) to assign ABO blood group (A2, B, and O) relative to the A1 reference allele: rs8176704 (G/A) to annotate A2, rs8176749 (C/T) to annotate B, and rs687289 (A/G) to annotate O. We detected ABO glycans in red blood cell-rich buffy coat samples using an adaptation of published ABO forensic dried blood spot techniques. Briefly, samples were diluted and applied to a nitrocellulose membrane, blotted with murine monoclonal anti-A (Immucor) or murine monoclonal anti-B (Immucor) antibodies followed by and streptavidin-conjugated donkey anti-mouse IgG HRP secondary (Jackson ImmunoResearch). Blots were developed, imaged using ImageQuant 350 (GE), and scored semi-quantitatively by two blinded independent observers relative to red blood cell-rich buffy coat reference panels from normal blood donors. Comparison of ABO imputed genotype to detected ABO glycans in red blood cell-rich buffy coat identified discrepancies in the predicted major ABO blood group assignment (i.e. blood group O, A, B, or AB) in 4.4% of individuals. Interestingly, A glycan density was generally low in individuals with a discrepant A phenotype, suggesting that these individuals harbor a less robust blood group A enzyme variant. There was no common pattern for genotype-phenotype discrepancies. Multiple types of errors (both presence and absence of ABO glycans) were observed, and divergent patterns of ABO glycans were observed between homozygotes for ABO SNVs, suggesting multiple ABO alleles or other genetic modifiers underlie the discordant calls. Interrogation of the Exome Variant Server (http://evs.gs.washington.edu/EVS/) found one-third of ABO nonsynonymous polymorphisms in the exome data set are novel, indicating a large proportion of the genetic variation at ABO has not been previously described. Furthermore, the majority of nonsynonymous variants in the EVS database for the genes encoding H antigen (the ABO acceptor structure), FUT1 and FUT2, were also novel. In summary, ABO(H) is more variable genetically than previously suspected. A commonly applied ABO SNV genotyping algorithm resulted in 4.4% frequency of major ABO blood group glycan discrepancy as detected in red blood cell-rich buffy coats, likely due in large part to the underlying genetic diversity at ABO. Further work is needed to deeply characterize the genetic diversity at ABO, FUT1, and FUT2 and correlate the spectrum of allelic variants with ABO(H) carbohydrate phenotypes. Such high resolution ABO and H classification would likely advance our understanding of the complex physiologic processes influenced by ABH blood groups, including disorders of hemostasis and thrombosis, vascular disease, infectious disease susceptibility, complications of pregnancy, and allogeneic exposures. Disclosures: Ginsburg: Shire Pharmaceuticals: Membership on an entity's Board of Directors or advisory committees; Portola Pharmaceuticals: Consultancy; Catalyst Biosciences: Consultancy; Baxter Pharmaceuticals: benefit from payments to Children's Hosptial, Boston, and the University of Michigan Patents & Royalties; Merck Pharmaceuticals: Consultancy.