Paratome: an online tool for systematic identification of antigen-binding regions in antibodies based on sequence or structure

Abstract

Antibodies are capable of specifically recognizing and binding antigens. Identification of the antigen-binding site, commonly dubbed paratope, is of high importance both for medical and biological applications. To date, the identification of antigen-binding regions (ABRs) relies on tools for the identification of complementarity-determining regions (CDRs). However, we have shown that up to 22% of the residues that actually bind the antigen fall outside the traditionally defined CDRs. The Paratome web server predicts the ABRs of an antibody, given its amino acid sequence or 3D structure. It is based on a set of consensus regions derived from a structural alignment of a non-redundant set of all known antibody-antigen complexes. Given a query sequence or structure, the server identifies the regions in the query antibody that correspond to the consensus ABRs. An independent set of antibody-antigen complexes was used to test the server and it was shown to correctly identify at least 94% of the antigen-binding residues. The Paratome web server is freely available at http://www.ofranlab.org/paratome/.

Figure 1.

An example HTML results file of running Paratome on anti IL-15 (PDB id 2xqb). (A) The sequences of Ab chains with ABRs highlighted. The location of ABRs residues is indicated according to the numbering in the ATOM field within the input PDB file. Note that for ABR L1, the location of five residues (S, N, L, K and R) is indicated with an asterisk as they do not appear in the ATOM field within the PDB file. A link to the visualization of the analyzed complex is located below the list of identified ABRs. (B) The visualization of the analyzed 2qxb complex using the Jmol Java applet.

Figure 2.

Antigen-binding regions identification. (A) The sequence of the query Ab is BLASTed against the non-redundant set of annotated Abs. (B) The framework regions (FRs) of the best BLAST hit Ab are aligned to the query Ab (sequences—BLAST, structures—CE). (C) The ABRs of the query Ab are inferred according to the ABRs of the best BLAST hit Ab.