The classification of the complementarity determining regions (CDRs) of antibodies presented on this website and the accompanying database is based on a new clustering of CDR conformations now available on biorxiv. In this work, we used the EDIA electron density metric (Meyder et al., 2017) to produce high-quality data sets eliminating CDR structures with low electron density support (due to low resolution, dynamics within the crystal, or misfitting). We used a distance function based on the maximum backbone dihedral angle difference across phi, psi, and omega to compare each pair of CDRs of the same length for a given CDR (H1, H2, etc.). Clustering was performed with the density based algorithm, DBSCAN, which removes outlier structures considered as noise. Clusters were established if they were present in high electron density data sets and contained at least 10 unique sequences across the whole PDB. The current database includes assignment of CDRs H1, H2, H3, H4, L1, L2, L3, and L4 in all antibody domains in the PDB to this new classification system (H4 and L4 are the “DE loop” that connects strands D and E, and sits adjacent to CDR1 and CDR2 in the light and heavy chain variable domains; it sometimes interacts with antigens and affects the conformations of CDR1 and/or CDR2). We also identify CDR structures that do not fit into any of the canonical clusters and these are labeled as “starred clusters” (e.g., H1-13-*). The database includes careful assignments of IMGT species and germlines for the V-region of all antibodies in the PDB, including the identification of non-human grafts of CDRs into predominantly human framework regions. The database can be searched by PDB code, cluster name (e.g. L1-11-1), and by IMGT germline. See the “Help Search” button for more information on searching. The “Help Submit” button provides information on uploading sequences of structures for analysis of the CDR sequences and conformations.

Please refer to our recent bioRxiv paper for more information.