Final Report Summary - IFLPD (Incorporating flexibility into protein-ligand docking)
Project context and objectives
Structure-based drug design (SBDD) is routinely used to identify new active compounds in early stages of drug development. Among the computer-aided drug design approaches, docking methods are the tools of choice for the prediction of ligand-receptor interactions. Unfortunately, the success of molecular docking is hampered by the sparse representation of protein dynamics. The principal objective of this project has been to develop methods and software to introduce protein flexibility in the context of small molecule docking and protein-protein docking. Many universities and pharmaceutical companies are using the public applications and websites created during the project.
Work performed
Dealing efficiently with protein flexibility and understanding the effects of dynamics in ligand binding is one of the main challenges in the world of computer-aided drug design. One simple and efficient way of representing protein plasticity is by using multiple 'static' receptor conformations, also known as ensemble docking. The ensembles can consist of multiple experimental structures, computationally created models, or both. Most of the results obtained during this project consist of variations of the ensemble docking procedure, applied in different medicinal chemistry scenarios. The most significant results are summarised below.
Exploiting experimental conformational diversity in protein-ligand docking: Experimental 3-D structures are the main source of conformational variability for protein-ligand docking and virtual screening campaigns. According to our studies, ensemble docking with X-ray protein structures displayed better discrimination of actives from inactive ligands than single receptor docking. Moreover, the use of multiple receptor conformations enhanced the chemical diversity of the compounds. On average, ensembles consisting of three to five conformers generated better discrimination values than individual conformations (see references 1-3).
%ALiBERO: This is a new computational method that improves the recognition of active compounds in distant homology models. For ~50 % of the therapeutically relevant protein targets, no experimental 3-D structures are available. For these cases, homology modelling can be used to generate protein models. Unfortunately, raw homology models usually display poor docking performance. During this project, we developed a method that optimises homology models so that their docking performance is on a par with X-ray structures. The method is called ALiBERO (3) (automatic ligand-guided backbone ensemble receptor optimisation) and performs an iterative search based on two main steps:
- the generation of multiple receptor conformers (4);
- a selection of the conformers according to docking/virtual screening (VLS) performance.
The method has been successfully applied to many important nervous system targets, such as the A2A adenosine receptor, the dopamine D3 receptor and the 5-HT1A serotonin receptor. We also developed a method to check the precision of the models (SimiCon) that allows automated identification of equivalent protein-ligand atomic contacts (5). SimiCon was implemented in a free web server located at http://abagyan.ucsd.edu/SimiCon
Worldwide modelling assessment on GPCR structure: A GPCR Dock 2010 community-wide assessment was conducted by our group to evaluate the status of molecular modelling and ligand docking for three recent GPCR targets of varying difficulty: dopamine D3 and CXCR4 chemokine receptors bound to small molecule antagonists and CXCR4 with a synthetic cyclopeptide. According to the results, the fact that D3 had a closer homolog for homology modelling, combined with the use of modern docking protocols and QSAR information, allowed for the prediction of complexes with atomic details of accuracy approaching experimental. On the contrary, CXCR4 complexes, which only possess distant homology to the available GPCR structures, still remain very challenging (6). The results were published on the website: http://abagyan.ucsd.edu/GPCRDock2010
Protein-protein interactions: In a recent paper (7), we investigated the cost of backbone conformational changes upon association in a large dataset consisting of 2 090 unique unbound-to-bound domain transitions. According to our estimates, 65 % of the transitions did not show significant changes upon binding (i.e. the RMSD unbound to bound was = 1.5 Å), 13 % explored the bound conformation in the unbound state (conformational selection model), and only 2 % clearly require external energy (induced fit model). An important fact that arose from the study was that domains with many partners tend to undergo smaller changes upon association and are less likely to freely explore larger adaptations. Some of the estimates were derived from molecular dynamic (MD) simulations performed to 70 protein domains, starting from the unbound form. The trajectories were part of a large repository of MD simulations (MoDEL: molecular dynamics extended library) which consists of > 1 700 trajectories of proteins representative of monomeric soluble structures in the protein data bank (PDB)(8).
References
1. Bottegoni G., Rocchia W., Rueda M., Abagyan R. and Cavalli A. (2011). 'Systematic exploitation of multiple receptor conformations for virtual ligand screening'. PLoS One 6(5):e18845. 2. Rueda M., Bottegoni G. and Abagyan R. (2010). 'Recipes for the Selection of Experimental Protein Conformations for Virtual Screening'. Journal of Chemical Information and Modeling 50(1):186-193. 3. Rueda M., Totrov M. and Abagyan R. (2012). 'ALiBERO: evolving a team of complementary pocket conformations rather than a single leader'. Journal of Chemical Information and Modeling 52(10):2705-2714. 4. Rueda M., Bottegoni G. and Abagyan R. (2009). 'Consistent improvement of cross-docking results using binding site ensembles generated with elastic network normal mode'. Journal of Chemical Information and Modeling 49(3):716-725. 5. Rueda M., Katritch V., Raush E. and Abagyan R. (2010). 'SimiCon: a web tool for protein-ligand model comparison through calculation of equivalent atomic contacts'. Bioinformatics 26(21):2784-2785. 6. Kufareva I., Rueda M., Katritch V., Stevens R.C. and Abagyan R. (2011). 'Status of GPCR modeling and docking as reflected by community-wide GPCR Dock 2010 assessment'. Structure 19(8):1108-1126. 7. Stein A., Rueda M., Panjkovich A., Orozco M. and Aloy P. (2011). 'A systematic study of the energetics involved in structural changes upon association and connectivity in protein interaction networks'. Structure 19(6):881-889. 8. Meyer T. et al. (2010). 'MoDEL (Molecular Dynamics Extended Library): a database of atomistic molecular dynamics trajectories'. Structure 18(11):1399-1409.
Structure-based drug design (SBDD) is routinely used to identify new active compounds in early stages of drug development. Among the computer-aided drug design approaches, docking methods are the tools of choice for the prediction of ligand-receptor interactions. Unfortunately, the success of molecular docking is hampered by the sparse representation of protein dynamics. The principal objective of this project has been to develop methods and software to introduce protein flexibility in the context of small molecule docking and protein-protein docking. Many universities and pharmaceutical companies are using the public applications and websites created during the project.
Work performed
Dealing efficiently with protein flexibility and understanding the effects of dynamics in ligand binding is one of the main challenges in the world of computer-aided drug design. One simple and efficient way of representing protein plasticity is by using multiple 'static' receptor conformations, also known as ensemble docking. The ensembles can consist of multiple experimental structures, computationally created models, or both. Most of the results obtained during this project consist of variations of the ensemble docking procedure, applied in different medicinal chemistry scenarios. The most significant results are summarised below.
Exploiting experimental conformational diversity in protein-ligand docking: Experimental 3-D structures are the main source of conformational variability for protein-ligand docking and virtual screening campaigns. According to our studies, ensemble docking with X-ray protein structures displayed better discrimination of actives from inactive ligands than single receptor docking. Moreover, the use of multiple receptor conformations enhanced the chemical diversity of the compounds. On average, ensembles consisting of three to five conformers generated better discrimination values than individual conformations (see references 1-3).
%ALiBERO: This is a new computational method that improves the recognition of active compounds in distant homology models. For ~50 % of the therapeutically relevant protein targets, no experimental 3-D structures are available. For these cases, homology modelling can be used to generate protein models. Unfortunately, raw homology models usually display poor docking performance. During this project, we developed a method that optimises homology models so that their docking performance is on a par with X-ray structures. The method is called ALiBERO (3) (automatic ligand-guided backbone ensemble receptor optimisation) and performs an iterative search based on two main steps:
- the generation of multiple receptor conformers (4);
- a selection of the conformers according to docking/virtual screening (VLS) performance.
The method has been successfully applied to many important nervous system targets, such as the A2A adenosine receptor, the dopamine D3 receptor and the 5-HT1A serotonin receptor. We also developed a method to check the precision of the models (SimiCon) that allows automated identification of equivalent protein-ligand atomic contacts (5). SimiCon was implemented in a free web server located at http://abagyan.ucsd.edu/SimiCon
Worldwide modelling assessment on GPCR structure: A GPCR Dock 2010 community-wide assessment was conducted by our group to evaluate the status of molecular modelling and ligand docking for three recent GPCR targets of varying difficulty: dopamine D3 and CXCR4 chemokine receptors bound to small molecule antagonists and CXCR4 with a synthetic cyclopeptide. According to the results, the fact that D3 had a closer homolog for homology modelling, combined with the use of modern docking protocols and QSAR information, allowed for the prediction of complexes with atomic details of accuracy approaching experimental. On the contrary, CXCR4 complexes, which only possess distant homology to the available GPCR structures, still remain very challenging (6). The results were published on the website: http://abagyan.ucsd.edu/GPCRDock2010
Protein-protein interactions: In a recent paper (7), we investigated the cost of backbone conformational changes upon association in a large dataset consisting of 2 090 unique unbound-to-bound domain transitions. According to our estimates, 65 % of the transitions did not show significant changes upon binding (i.e. the RMSD unbound to bound was = 1.5 Å), 13 % explored the bound conformation in the unbound state (conformational selection model), and only 2 % clearly require external energy (induced fit model). An important fact that arose from the study was that domains with many partners tend to undergo smaller changes upon association and are less likely to freely explore larger adaptations. Some of the estimates were derived from molecular dynamic (MD) simulations performed to 70 protein domains, starting from the unbound form. The trajectories were part of a large repository of MD simulations (MoDEL: molecular dynamics extended library) which consists of > 1 700 trajectories of proteins representative of monomeric soluble structures in the protein data bank (PDB)(8).
References
1. Bottegoni G., Rocchia W., Rueda M., Abagyan R. and Cavalli A. (2011). 'Systematic exploitation of multiple receptor conformations for virtual ligand screening'. PLoS One 6(5):e18845. 2. Rueda M., Bottegoni G. and Abagyan R. (2010). 'Recipes for the Selection of Experimental Protein Conformations for Virtual Screening'. Journal of Chemical Information and Modeling 50(1):186-193. 3. Rueda M., Totrov M. and Abagyan R. (2012). 'ALiBERO: evolving a team of complementary pocket conformations rather than a single leader'. Journal of Chemical Information and Modeling 52(10):2705-2714. 4. Rueda M., Bottegoni G. and Abagyan R. (2009). 'Consistent improvement of cross-docking results using binding site ensembles generated with elastic network normal mode'. Journal of Chemical Information and Modeling 49(3):716-725. 5. Rueda M., Katritch V., Raush E. and Abagyan R. (2010). 'SimiCon: a web tool for protein-ligand model comparison through calculation of equivalent atomic contacts'. Bioinformatics 26(21):2784-2785. 6. Kufareva I., Rueda M., Katritch V., Stevens R.C. and Abagyan R. (2011). 'Status of GPCR modeling and docking as reflected by community-wide GPCR Dock 2010 assessment'. Structure 19(8):1108-1126. 7. Stein A., Rueda M., Panjkovich A., Orozco M. and Aloy P. (2011). 'A systematic study of the energetics involved in structural changes upon association and connectivity in protein interaction networks'. Structure 19(6):881-889. 8. Meyer T. et al. (2010). 'MoDEL (Molecular Dynamics Extended Library): a database of atomistic molecular dynamics trajectories'. Structure 18(11):1399-1409.
