MOTIVATION: Antibodies are able to recognize a wide range of antigens through their complementary determining regions formed by six hypervariable loops. Predicting the 3D structure of these loops is essential for the analysis and reengineering of novel antibodies with enhanced affinity and specificity. The canonical structure model allows high accuracy prediction for five of the loops. The third loop of the heavy chain, H3, is the hardest to predict because of its diversity in structure, length and sequence composition. RESULTS: We describe a method, based on the Random Forest automatic learning technique, to select structural templates for H3 loops among a dataset of candidates. These can be used to predict the structure of the loop with a higher accuracy than that achieved by any of the presently available methods. The method also has the advantage of being extremely fast and returning a reliable estimate of the model quality. AVAILABILITY AND IMPLEMENTATION: The source code is freely available at http://www.biocomputing.it/H3Loopred/ .
Bibliographical noteKAUST Repository Item: Exported on 2020-10-01
Acknowledged KAUST grant number(s): KUK-I1-012-43
Acknowledgements: Funding: KAUST Award No. KUK-I1-012-43 made by King Abdullah University of Science and Technology (KAUST).
This publication acknowledges KAUST support, but has no KAUST affiliated authors.