INVESTIGADORES
PETRUCCELLI Silvana
congresos y reuniones científicas
Título:
Structural modeling of an anti-transglutaminase antibody and analysis of its binding mode to transglutaminase using blind rigid-body protein-protein docking
Autor/es:
VECCHI, BRUNO; PETRUCCELLI, SILVANA
Lugar:
Quilmes
Reunión:
Congreso; 1er Congreso Argentino de Bioinformática y Biología Computacional (CAB2C; 2010
Institución organizadora:
Sociedad Argentina de Bioinformatica
Resumen:
Human tissue transglutaminase (htTG) is an enzyme that plays an important role in blood coagulation. Incidentially, it is the main auto-antigen of Celiac disease patients; therefore serum reactivity against it is one of the main indicators of Celiac disease[6]. We cloned and sequenced a murine monoclonal antibody against htTG. In order to gain knowledge regarding the antibody-antigen binding mode and the nature of the contact surface, we proceeded to perform molecular modeling of its variable domains, followed by blind, rigid-body protein-protein docking against the antigen?s crystallized protein structure. Modeling The modeling of the variable zones of the antibody heterodimer was done by homology modeling for the structurally conserved zones. For the complementary determining regions (CDRs) that define the antibody specificity (and for which there are almost never reliable templates to derive the structure from), two independent approaches were carried out: ab initio loop refinement using Modeller[4], and knowledge-based assessment of loop conformations using the canonical structures[7] method as implemented by Rosetta[2]. Rigid body protein protein docking These structures were used to perform rigid body docking simulations against the crystallized htTG structure (PDBID 1KV3A) using ZDOCK[1]. Results show that, for both modeling methods, there is a clear energy funnel in score vs rmsd plots that show a consensual and highly favored binding mode, suggesting binding specificity. In contrast, two negative control antibodies not only yielded lower binding energy scores, but also top-scoring decoys lacked similarity, as judged by their large mutual rmsd values. Clustering and residue-residue contact map Average residue-residue contact maps were calculated. First, all 50.000 decoys of each model were clustered into groups using the k-medoids clustering algorithm. A contact map for each member of the top 2 scoring clusters was done by means of Voronoi?s polyhedra method[5] and averaged within each cluster. The resulting maps show that there is a high similarity among the contact surfaces of different m odels. Moreover, the antigen binding zone corresponds very closely with the antibody?s epitope, as previously determined using phage display[3]. This confirms that both the modeling and docking procedures faithfully represent their molecular entities and that further residue-residue contact information can be extracted from this models, which may guide the design of potential single or multi-point mutants with increased affinity.