Flexible Docking of Biotin into Streptavidin

Pharmaceutical case study Page 1 of 5 Flexible Docking of Biotin into Streptavidin -an Affinity Validation Study Key Products Insight11 Affinity Discover-3 Binding Site Analysis Ludi DeCipher Accelrys European Headquarters 334 Cambridge Science Park Cambridge, CB4 0WN, UK Tel: +44 1223 228500 Accelrys Asia Headquarters Nishi-shimbashi TS Bldg 11F Nishi-shimbashi 3-3-1, Minato-ku, Tokyo, 105-0003, Japan Tel: 81 3 3578 3861 Accelrys Corporate Headquarters 9685 Scranton Road San Diego, CA 92121-3752, USA Tel: +1 858 799 5000 Dr Teresa Lyons, Accelrys, has used Affinity and supporting tools within InsightII to reproduce the crystal structure of streptavidin complexed with biotin (PDB ID 1STP). This complex is one of several being used to validate Affinity as a docking application and to develop a protocol to successfully differenntiat the correct complex structure from the set of structures Affinity produces. See also: • Flexible Docking of 3-Phenylpropylamine into Trypsin -an Affinity Validation Study (www.accelrys.com/cases/affinity/affinity1.pdf) • Flexible Docking of Benzylsuccinate Inhibitor into Carboxypeptidase A -an Affinity Validation Study (www.accelrys.com/cases/affinity/affinity2.pdf) The computational challenge of docking ligands into receptors involves two distinct steps: • Finding the correct conformation and orientation of the ligand in the bindiin pocket • Correctly identifying that particular ligand pose from a set of energetically reasonable conformations and orientations. This is widely known as the docking and scoring problem. Affinity distinguishes itself from other docking programs in that it allows both the ligand and the binding pocket to be flexible during the calculation. Two methods can be used for the docking -grid or simulated annealing. Grid docking is generally quicker (and less accurate) because the energetics of the binding pocket are predefined using a grid whereas simulated annealing docking uses full forcefield energetics to optimize the position of the ligand in the binding pocket. Industry Sector Pharmaceuticals Organization AccelrysPharmaceutical case study continued Page 2 of 5 Affinity starts by rotating, translating, and doing a conformational search of the ligand in the context of the binding pocket using a Monte Carlo search algorithm. This is optionally followed by a Discover-based minimization or dynamics phase that allows both the ligand and the binding pocket to optimiiz the interactions. An Affinity run results in a set of structures that fall within the energetic and geometric criteria set by the user. Assuming that conformational space has been searched adequately during the Affinity run, the challenge then becomes how to effectively pick out 'the real structure' from the result set. Methods • Defining the binding pocket: In order to be as unbiased as possible, the binding pocket was defined based on an 'Active Site Search', a cavity-findiin algorithm that is part of the Binding Site Analysis module of InsightII. Whole residues within 3 Angstroms of the cavity surface were collected to make up the binding site subset. For carboxypeptidase A, the largest cavity found coincided well with the binding pocket where benzylsuccinate sits in the crystal structure. • Setting up the Affinity job: The Grid Docking Affinity job was run using the default parameters for the grid setup. 50 structures were collected. • Minimization: All parameters were left at their default values except the Energy Range was set to 200 and Energy Tolerance to 1e+6. • Dynamics: The dynamics stage was run using the default parameters. • Analysis of the structures: The resulting set of structures were evaluated for how close they were to the crystal structure 1STP based on an all-atom RMSD of biotin only (a function set up in Decipher). Based on this criterioon the structure most similar to the crystal structure was frame 41 of the trajectory file. All frames were analyzed using four methods: • Total energy as output by Affinity • Interaction energy between the ligand and protein as calculated with the Evaluate/Intermolecular command (Docking module) • Ludi Score 3 of the interaction (Ludi module) • Number of hydrogen bonds between protein and ligand. A BCL script facilitated the batch analysis of the trajectory file. 􀁓 Figure 1 Superimposition of Frame 5 of the Affinity run (best ranked overall) to the crystal structuur 1STP. The ball-and-stick rendeere ligand is from the crystal structure; the stick rendered ligand is frame 5. The surrounding orange residues constitute the binding pocket (flexible residues during the Affinity run). Other protein atoms are in purple. Pharmaceutical case study continued Page 3 of 5 Results Table 1 shows the results of the analysis, highlighting in dark yelllo the RMSDs that are within 1.5 Angstroms of the crystal structure (ligand to ligand). Also highlighted in dark yellow are the top 5 ranked frames for each analysis category and in lighter yellow the next top 5 (ranks 6 through 10). In bold italics are the frames that fall into the top 20% of 3 out of 4 of the analysis categories. Frame RMSD TotalE InterE Ludi3 Hbonds 1 0.693 -148.61 -67.48 504 6 2 1.182 -363.08 -62.33 342 3 3 0.876 -530.08 -59.41 364 3 4 1.201 -530.96 -52.92 378 2 5 1.429 -536.75 -53.02 392 5 6 6.552 -506.51 -44.01 382 7 7 6.473 -493.23 -30.64 283 5 8 6.093 -486.70 -26.97 196 0 9 5.581 -496.77 -22.14 159 0 10 6.082 -499.78 -24.24 165 0 11 6.104 -479.59 -29.44 252 4 12 6.449 -513.66 -37.15 201 2 13 4.750 -518.10 -30.96 144 0 14 6.153 -513.50 -31.02 176 2 15 5.869 -518.97 -35.99 181 3 16 4.624 -443.19 -6.28 268 5 17 6.337 -518.24 -38.33 247 2 18 6.119 -512.09 -33.24 194 1 19 6.446 -500.37 -30.81 292 2 20 6.403 -496.98 -23.60 202 1 21 5.185 -518.63 -28.59 186 1 22 6.602 -529.26 -38.94 202 3 23 5.004 -526.14 -29..33 164 1 24 4.085 -528.15 -31.06 177 1 25 4.092 -529.44 -38.33 184 1 26 7.054 -462.09 -21.60 211 7 27 5.989 -500.42 -30.78 277 6 28 5.842 -520.40 -36.42 198 3 29 5.893 -532.76 -42.18 290 4 30 5.757 -529.81 -38.10 156 1 31 4.780 -500.64 -32.24 199 2 32 4.926 -534.91 -45.52 222 1 33 5.045 -540.44 -48.53 221 1 34 6.087 -536.60 -43.19 216 0 35 6.003 -521.14 -32.49 170 2 36 4.776 -459.32 -28.00 256 3 37 5.910 -514.01 -39.52 188 1 38 6.316 -505.91 -28.70 139 0 39 6.644 -500.62 -24.59 144 0 40 7.103 -508.69 -28.16 140 0 41 2.854 -524.25 -41.32 227 0 42 6.508 -507.47 -30.65 219 1 43 2.208 -532.94 -52.01 328 2 44 4.261 -440.92 -29.91 355 5 45 3.631 406.62 -9.36 208 2 46 2.992 -489.16 -30.14 2.64 4 47 6.023 -505.35 -19.72 136 0 48 6.741 -520.26 -34.89 143 0 49 6.173 -533.19 -45.36 222 1 50 6.315 -535.22 -42.42 216 1 􀁓Pharmaceuticals case study continued Page 4 of 5 Conclusions The structures thatmost closely match the crystal structure are frames 1,3,2,4, and 5 (in order of lowest RMSD to highest). In the analysis of the data, frames 3, 4, and 5 all ranked in the top ten (top 20%) for total energy, interaction energgy and Ludi score. Frame 43, with an RMSD of 2.20 Angstroms, is the only other frame to also be in the top ten using these three analysis methods. If the set of structures is further culled based on the number of hydrogen bonds formed between ligand and receptor, frame 5 would be the structure of choice, ranking third with 5 out of 7 hydrogen bonds. If the ranking is taken down to the top 10% (or top 5 of this set of structures), only frame 5 is consistently at the top of the analysis for the three primary analysis methods (i.e. not including hydrogen bonds). Interestingly, the complle CLOSEST to the crystal structure, frame 1, doesn't rank in the top 10 for total energy. In this case, the analysis finds ligand poses that are within 1.5 Angstroms RMSD of the crystal structure but does not find the top ranked structure (frame 1) as determined by RMSD. Flexible Docking of Biotin into Streptavidin -an Affinity Validation Study