Predicting the structure of the antibody variable region from sequence has been the focus of considerable research since the work of Kabat and Wu (1), Padlan et al. (2), Stanford and Wu (3), and Feldmann et al. (4). Following the essentially “homology”-based predictions of these early approaches, methods were developed that introduced more rule-based procedures exemplified by our own work (5–11) and that of Chothia, Lesk, and colleagues (12–17) who, building on the observations of Kabat et al. (18,19) and Padlan and Davies (20), developed the concept of canonical classes for certain of the variable region complementarity-determining region (CDRs). Since then, attention has focused on the more difficult problem of non-canonical CDRs, of which the antibody heavy-chain CDR3 (hereafter referred to as H3) is the most unique.