Antisense oligonucleotides, which act through the pairing of complementary bases to an RNA target sequence, are showing great promise in research and clinical applications. However, the selection of effective antisense oligonucleotides has proven more difficult than initially presumed. We developed a prediction algorithm to identify those sequences with the highest predicted binding affinity for their target mRNA based on a thermodynamic cycle that accounts for the energetics of structural alterations in both the target mRNA and the oligonucleotide. The model was used to predict the binding affinity of antisense oligonucleotides complementary to the rabbit β€-globin (RBG) and mouse tumor necrosis factor-β£ (TNFβ£) mRNAs, for which large experimental datasets were available. Of the top ten candidates identified by the algorithm for the RBG mRNA, six were the most strongly binding sequences determined from an experimental assay. The prediction for the TNFβ£ mRNA also identified high affinity sequences with βΌ60% accuracy. Computational prediction of antisense efficacy is more cost-efficient and faster than in vitro or in vivo selection and can potentially speed the development of sequences for both research and clinical applications.