Table S3.Methods used for HSP90 pose predictions. The Method ID corresponds to the numbers in the in-text figures, the Submission ID is a unique ID given to each prediction upon submission, the Participant’s Name is the last and first name of the participant, the Method Name is the informal name given to the procedure usedfor pose prediction andMethod Descriptions are condensed from submitted protocol descriptions.
Method ID / Submission ID / Participant Name / Method Name / Method Description / Median RMSD (Å) Rank 1 / Median RMSD (Å) Best of Top 51 / 564faa741984e / Ye, Zhaofeng / Vina-Smina / Found & docked receptors with most similar HSP90 compounds; MD and clustering of conformations; Manual check on poses. DOI: 10.1007/s10822-016-9941-0 / 0.34 / 0.28
2 / 564e43759677e / Moro, Stefano / Gold-PlantsPLP-rDOCK / Benchmarked 13 ligand-protein complexes with DockBench 1.01 used to determine best protocol; Visually inspected poses and evaluated poses by MD monitoring the DSF (dynamic scoring function); Best DSF selected. See: JCAMD 2016 D3R-GC Special Issue / 0.48 / 0.48
3 / 564ee243b0bf4 / Rognan, Didier / Surflex-GRIM / Docking to all 4 receptors with 0-2 binding site waters with Surflex-Dock (20 poses); rescoring by interaction pattern graph matching (GRIM) to 176 PDB templates; best 5 GRIM scores. DOI: 10.1007/s10822-016-9930-3 / 0.54 / 0.54
4 / 564d175b04f55 / Kumar, Ashutosh / RosettaLigand-Omega-ROCS / PDB ligand conformation with highest 3D shape similarity docked and refined; Structures with highly shape similar crystal ligand used; Sorted 200 poses; Best “interface delta” score selected. Cite this article as: Kumar, A. & Zhang, K.Y.J. J Comput Aided Mol Des (2016). doi:10.1007/s10822-016-9931-2 / 0.62 / 0.51
5 / 5665854ce7cf9 / Grudinin, Sergei / Vina-ConvexPL / Manually selected receptors; RL pair used and docked without flexible sidechains; Open conformation of aminopyrimidine receptor used; Re-scored and best 5 poses reported. See: JCAMD 2016 D3R-GC Special Issue / 0.64 / 0.49
6 / 564f4994c90c9 / Michel, Julien / rDOCK / Assign crystallographic structures on PDB to 3 clusters according to RMSD of sidechains residues (binding site); Added H; Visually inspected 15 highest ranked and picked 5. / 0.65 / 0.65
7 / 564e50f3b179d / Michel, Julien / rDOCK / Variant of #6. / 0.67 / 0.63
8 / 564f897715791 / Zou, Xiaoqin / Vina-ITScore / Collected PL complex structures from PDB; Calculated and used highest structure similarity with known binders; Cluster cutoff 0.5A; 500 models evaluated with ITScore. / 0.67 / 0.58
9 / 564f7dbd16a13 / Koes, David / Vina-Smina / Docked to all 4 receptors with 0-2 binding site waters; Top pose selected; Training set balanced (#active=#decoy). DOI: 10.1007/s10822-016-9960-xManuscript 16-00166 / 0.75 / 0.75
10 / 564f722a7f45a / Anonymous / Vina-MMPBSA / Docked in all 4 receptors with binding site waters. Pose with best scoring submitted. / 0.76 / 0.76
11 / 564fb9530127a / Iorga, Bogdan / Gold / 6 ligands docked with K58 as flexible residue; Superimposed PL complexes, calculated common center of mass, and defined 3 ref ligands; Manually examined and picked poses based on ability to make interaction with residues of active site. See: JCAMD 2016 D3R-GC Special Issue / 0.90 / 0.90
12 / 564ca5a11a3d7 / Cho, Art / Glide-Prime-Desmond-Qsite / No details. / 0.91 / 0.62
13 / 564f443622298 / Santos Martins, Diogo / Vina / Added H; 2 receptor models used; 1UYG penalized to compensate for systematically lower energy obtained; Penalty value decided upon visual inspection; Top 5 lowest energy submitted. DOI: 10.1007/s10822-016-9943-y / 0.98 / 0.90
14 / 564e510834315 / Michel, Julien / rDOCK / Variant of #6 & #7; Top 5 ranked poses using 2CCU. / 1.03 / 1.03
15 / 564e510347f99 / Michel, Julien / rDOCK / Variant of #6 & #7; Top 5 ranked poses using 4CWF. / 1.10 / 1.10
16 / 564fb613b0cce / Rauscher, Fabian / Surflex / Followed gdock protocol from Surflex-Dock manual. DOI: 10.1007/s10822-015-9846-3. / 1.29 / 1.29
17 / 564f7b05c71e5 / Ruiz Carmona, Sergio / rDOCK / No details. See: JCAMD 2016 D3R-GC Special Issue / 1.40 / 1.40
18 / 564e50f6634a5 / Michel, Julien / rDOCK / Variant of #6 & #7; Ranked according to docking energy. / 1.65 / 0.67
19 / 564e50fb3ccb3 / Michel, Julien / rDOCK / Variant of #6 & #7; Top 5 ranked using 4W7T. / 1.65 / 1.06
20 / 564e5106206c5 / Michel, Julien / rDOCK / Variant of #6 & #7; Top 5 ranked using 2FWZ. / 1.86 / 0.95
21 / 564e50ff31d3e / Michel, Julien / rDOCK / Variant of #6 & #7; Top 5 ranked using 4L94. / 2.05 / 1.11
22 / 564f81fb80465 / Koes, David / Vina-Smina / Docked to all 4 receptors with 0-2 binding site waters; Selected top poses using neural net (20 nodes & 2 output nodes); Training set balanced (#active=#decoy). DOI: 10.1007/s10822-016-9960-x / 2.25 / 2.25
23 / 564fa6bf8c79b / Bansal, Nupur / Glide SP-Mtflex / Protonation at pH 7; H optimized and heavy atoms fixed; Capped N & C termini of protein; Retained crystal waters; Docked 6 ligands in 4 protein conformations; Top 5 scoring ligands retained / 2.45 / 1.92
24 / 564fa6bf8c79b_1 / Bansal, Nupur / Glide SP-Mtflex / Variant of #23 / 2.45 / 1.92
25 / 564fbfb26cb6b / Fu, Darwin / RosettaLigand-MOE-BCL. / Ligand conformational libraries generated with a combination of MOE and the in-house BioChemicalLibrary (BCL); Energetically minimized starting structure without altering protein backbone; 500 models made for each PL pair; Best scoring model against PL crystal submitted; 4 more selected for diversity / 3.28 / 2.05
26 / 564ec3204a49e / Ding, Feng / MedusaDock / Used all 4 backbone conformation; grouped similar poses by clustering analysis and ranked by “free-energy” based estimation; Selected a pose from each of top 5 clusters according to minimum energy / 3.39 / 2.17
27 / 564ec888d4faf / Park, Hwangseo / AutoDock-SYBYL / No details. / 3.73 / 2.17
28 / 564f823da563c / Koes, David / Vina-Smina / Docked to all 4 receptors with 0-2 binding site waters; Top pose selected. DOI: 10.1007/s10822-016-9960-x / 5.07 / 5.07
29 / 564f8224dfefd / Koes, David / Vina-Smina / Docked to all 4 receptors with 0-2 binding site waters; Selected top poses using neural net (20 nodes & 2 output nodes); Training set balanced (applied weight to decoys). DOI: 10.1007/s10822-016-9960-x / 5.08 / 5.08
30 / 564ad0860684b / Chung, Hwan Won / AutoDock-VnsDock / No details. / 5.29 / 5.08
31 / 564f82588372c / Koes, David / Vina-Smina / Docked to all 4 receptors with 0-2 binding site waters; Selected top poses using neural net (20 nodes & 2 output nodes); Training set balanced (#active=#decoy by downsampling). DOI: 10.1007/s10822-016-9960-x / 5.45 / 5.45
32 / 564f482577338 / Piotto, Stefano / Vina-YADA / YADA calculates best poses (Vina energy) and ranks based on several parameters and the distances between each ligand atom with the most conserved residues. Data is combined into a ranking function obtained by a Genetic Algorithm analysis. See: JCAMD 2016 D3R-GC Special Issue / 5.48 / 5.48
33 / 564fb8e4e36dd / Minh, David / Vina / Water box added with buffer of 10A; H removed from L; Hierarchical clustering with complete linkage based on ligand RMSDs; Pose (lowest interaction energy) ranked by population in the fully-bound state / 5.49 / 2.40
34 / 564fb99101321 / Minh, David / Vina / Variant of #32; Poses ranked by lowest observed interaction energy. / 5.49 / 2.16
35 / 564f4a42d3ee0 / Michel, Julien / Vina / Same as #6 & #7; Best Vina-generated poses for 2CCU are presented for each compound. / 5.56 / 0.63
36 / 564f81fc8469f / Koes, David / Vina-Smina / Variant of #22. DOI: 10.1007/s10822-016-9960-x / 5.65 / 5.65
37 / 564fb205838ab / Bansal, Nupur / Glide SP-Mtflex / Crystal waters retained in binding pocket area; Introduced sidechain flexibility while keeping backbone fixed; optimized H while heavy atoms are kept fixed; Capped N- and C-termini. / 5.66 / 4.22
38 / 564fb205838ab_1 / Bansal, Nupur / Glide SP-Mtflex / Variant of #37 / 5.66 / 4.22
39 / 564fb2075c714 / Bansal, Nupur / Glide SP-Mtflex / Glide SP docked the 6 ligands into protein conformers generated using Mtflex; Scoring done using movable type binding free energy method. / 6.63 / 6.45
Table S4. Methods used for MAP4K4 pose predictions.The Method ID corresponds to the numbers in the in-text figures, the Submission ID is a unique ID given to each prediction upon submission, the Participant’s Name is the last and first name of the participant, the Method Name is the informal name given to the procedure used andMethod Descriptions are condensed from submitted protocol descriptions.
Method ID / Submission ID / Participant Name / Method Name / Method Description / Median RMSD (Å) Rank 1 / Median RMSD (Å) Best of Top 51 / 566ff18cce846 / Cho, Art / Glide SP-Qsite / Glide 6.6 and QSite 6.6 of Schrodinger (that’s all the details given) / 1.64 / 1.64
2 / 5671f1dac24a1 / Moro, Stefano / Gold-PlantsPLP / DockBench 1.01 used to determine best protocol. Each ligand was docked using 2 selected protocols (GOLD goldscore and Plants-Plp) and rescored with MOE using GBVI/WSA. See: JCAMD 2016 D3R-GC Special Issue / 2.13 / 2.13
3 / 56717b1bf22b8 / Rauscher, Fabian / Surflex-Dock / Followed the gdock protocol from the Surflex-Dock manual and the process described
in J Comput Aided Mol Des. 2015; 29(6): 485–509. / 2.17 / 2.17
4 / 5671a605c7215 / Ye, Zhaofeng / Vina / The most similar co-crystal ligands were searched; docked the MAP ligands into the most similar co-crystal receptor; aligned the MAP compounds with the most similar co-crystal ligand and minimized with corresponding receptor; manually inspect poses. DOI: 10.1007/s10822-016-9941-0 / 2.35 / 1.98
5 / 567053ef92815 / Rognan, Didier / Surflex-GRIM / Docking to 8 PDB structures with Surflex-Dock (20 poses); filtering for h-bond to hinge; rescoring by interaction pattern graph matching (GRIM) to 8 PDB templates ; best 5 GRIM scores. DOI: 10.1007/s10822-016-9930-3 / 2.85 / 2.30
6 / 566f9ac64f4e9 / Kumar, Ashutosh / RosettaLigand-Omega-ROCS / PDB ligand conformation with highest 3D shape similarity docked and refined; Sorted 200 poses; Best “interface delta” score selected. Cite this article as: Kumar, A. & Zhang, K.Y.J. J Comput Aided Mol Des (2016). doi:10.1007/s10822-016-9931-2 / 2.87 / 1.62
7 / 5671a885d8be7 / Ruiz Carmona, Sergio / Docking-Undocking combination / No details / 2.88 / 2.58
8 / 5670cff292802 / Anonymous / Glide SP / Added hydrogens and assumed pH 7.4; All conformation for each ligand were used as a starting point for docking. / 3.15 / 3.15
9 / 5670edaea4508 / Ding, Feng / Medusa Dock / Used all 4 backbone conformations; Clustered and ranked poses based on “free-energy” based estimation; Selected representative pose for each top 5 clusters (min. energy). / 3.24 / 2.34
10 / 5671e4aa3df40 / Iorga, Bogdan / MOE-Gold / Superimposed, defined common center of mass of the binding site; LYS54 as flexible residue; Removed waters; Docked 30 ligands on each of 8 proteins with 200% precision; Filtered and manually selected final poses (most pertinent interactions with the binding site). See: JCAMD 2016 D3R-GC Special Issue / 3.94 / 3.79
11 / 56739a627cd8a / Bansal, Nupur / Glide SP / 1 water for 40B0 & 2 waters for 4U44 for docking; Included P-loop region; MT method used to calculate binding affinities; Top 5 scoring ligands retained. / 4.14 / 3.18
12 / 56739aac8685e / BansalNupur / Glide SP / Same as #8. / 4.55 / 2.41
13 / 56723485d93a5 / Michel, Julien / rDock / Visually inspected PDB structures & separated into 2 clusters (4rvt & 4u44) based on p-loop conformation; Imposed a pharmacophoric constraint consistent on 2 H-bond interactions to be established with the hinge region; Top 5 ranked poses docked on both receptors retained. / 4.56 / 4.20
14 / 5672349b60cdb / Michel, Julien / rDock / Same as #10. Top 5 ranked poses docked on 4rvt retained. / 4.69 / 3.55
15 / 567234bae8ead / Michel, Julien / rDock / Same as #10. Top 5 ranked poses docked on 4u44 retained. / 4.87 / 4.32
16 / 567091083fadd / Zou, Xiaoqin / Vina-ITScore / High resolution PDB structures used in structure similarity calculation were docked; 500 models generated; Re-docked with D3R conformation if failed to bind; Re-scored with ITScore. / 4.96 / 2.64
17 / 5671219a5ffbb / Liu, Zhihong / CDOCKER / Protein prepared with CHARMmforcefield, pH 7.4, ionic strength 0.145; Ligand binding site radius 10 A; Ionization method used to change ionization state of ligands. / 5.26 / 5.26
18 / 56739c2555358 / Bansal, Nupur / Glide SP-MtFlex / In-house code (moved backbone & side chain of binding pocket of R); Flexible backbone; Docked all 30 ligands in all MTflexb generated conformations; Top 5 scoring ligands retained base on the score. / 5.27 / 3.91
19 / 5671107cd7f46 / Park, Hwangseo / AutoDock-Sybyl / No details. / 5.33 / 2.41
20 / 567193f9c8bad / Koes, David / Vina-Smina / Docked to all 4 receptors with 0-2 waters; Top pose selected using training set that isn’t balanced; Custom scoring capabilities & calculated surface area terms. DOI: 10.1007/s10822-016-9960-x / 5.79 / 5.79
21 / 56739ae990476 / Bansal, Nupur / Glide SP-MtFlex / Backbone kept fixed; Applied Mtflex & Van der waal restraints to binding pockets; Waters retained for docking; Used SP to dock 6 ligands in all Mtflex generated conformations; Top 5 scoring ligands retained based on score. / 5.80 / 2.69
22 / 567193d5878ca / Koes, David / Vina-Smina / Variant of #17; Top pose selected using linear regression model; Training set balanced (#active=#decoy by downsampling decoys). DOI: 10.1007/s10822-016-9960-x / 5.95 / 5.95
23 / 567194475df58 / Koes, David / Vina-Smina / Variant of #17; DOI: 10.1007/s10822-016-9960-x / 6.27 / 6.27
24 / 56719413dc96d / Koes, David / Vina-Smina / Variant of #17; Used neural net (1 hidden layer, 20 nodes, 2 output nodes); Weights applied to decoys to balance without downsampling. DOI: 10.1007/s10822-016-9960-x / 6.38 / 6.38
25 / 5673820a92362 / Skwark, Marcin / Rosettaligand / Used additional 152 protein chains of closely homologous protein for docking; Pose with lowest Rosetta energy were clustered and 5 conformations with lowest normalized energy were chosen. / 6.5 / 5.66
26 / 567197718362c / Koes, David / Vina-Smina / Variant of #17; Top pose selected using LASSO regression model; Training set balanced (#active=#decoy by downsampling decoys). DOI: 10.1007/s10822-016-9960-x / 6.71 / 6.71
27 / 5671958e262bb / Grudinin, Sergei / Vina-ConvexPL / No details. See: JCAMD 2016 D3R-GC Special Issue / 6.87 / 3.81
28 / 5673816a0ae73 / Skwark, Marcin / Rosettaligand / Variant #22; 5 conformations with lowest energy were chosen among those docked to 4U44. / 7.52 / 3.91
29 / 56739b211906a / Bansal, Nupur / Glide SP-MtFlex / Variant of #18. / 7.65 / 3.68
30 / 56738138d0826 / Skwark, Marcin / Rosettaligand / Variant of #22; 5 conformations with lowest energy were chosen among those docked to 40B0. / 8.78 / 4.74
Table S5. Methods used for Stage 1 HSP90ligand ranking.The Method ID corresponds to the numbers in the in-text figures, the Submission ID is a unique ID given to each prediction upon submission, the Participant’s Name is the last and first name of the participant, the Method Name is the informal name given to the procedure used while Method Descriptions are condensed from submitted protocol descriptions. For submission files with multiple variants of the same method, only the highest score from each such submission is reported; the names of the methods for such submissions have number of prediction sets in the method name (parenthesis) and the scores range in the scores (parenthesis).
Method ID / Submission ID / Participant Name / Method Name / Method Description / Kendall’s Tau / Kendall Tau Err / Spearman Value / Spearman Err1 / 564faadc08579 / Ye, Zhaofeng / Vina-Smina (7) / Omega2 generated conformers for hsp compounds; align all hsp compounds to their most similar co-crystal ligands; minimize the aligned-conformers then score. DOI: 10.1007/s10822-016-9941-0 / 0.32
(0.24 - 0.32) / 0.047 / 0.45 / 0.06
2 / 564e50f6634a5 / Michel, Julien / rDOCK (2) / Assign crystal structures to 3 clusters based on RMSD of sidechains of residues in binding site; 2CCU, 4L94, 4CWF, 2FWZ and 4W7T selected; ranked by best scored pose obtained in all 5 structures. / 0.31
(0.29 - 0.31) / 0.051 / 0.44 / 0.07
3 / 565350f56d4c1 / Prathipati, Philip / QSAR (4) / QSAR model based affinity predictions using Random Forest regression. See: JCAMD 2016 D3R-GC Special Issue / 0.30
(0.25 - 0.30) / 0.048 / 0.43 / 0.07
4 / 564e50fb3ccb3 / Michel, Julien / rDOCK-d / Assign crystal structures to 3 clusters based on RMSD of sidechains of residues in binding site; Ranked by top ranked pose score obtained using 4W7T. / 0.25 / 0.053 / 0.36 / 0.07
5 / 564fb30adf44e / Submission retracted / 0.24 / 0.052 / 0.33 / 0.07
6 / 564f7d7c93745 / Koes, David / Vina-Smina (6) / Docked into the 4 receptors with 0-2 waters incl, top poses were selected using a LASSO regression model trained to distinguish actives/inactives. DOI: 10.1007/s10822-016-9960-x / 0.23
(-0.07 - 0.23) / 0.048 / 0.33 / 0.07
7 / 564e510347f99 / Michel, Julian / rDOCK-a / Assign crystal structures to 3 clusters based on RMSD of sidechains of residues for binding site; Ranked by top ranked pose score obtained using 4CWF. / 0.23 / 0.054 / 0.32 / 0.08
8 / 564f75fbeaa9e / Zou, Xiaoqin / ITScore (3) / ITScore-1; calibrated on 178 HSP90 complexes using the iterative method. / 0.22
(0.18 – 0.22) / 0.055 / 0.31 / 0.08
9 / 564f897715791 / Zou, Xiaoqin / ITScore / Variant of #6 / 0.22 / 0.055 / 0.31 / 0.08
10 / 564e5106206c5 / Cho, Art / rDOCK-b / Assign crystal structures to 3 clusters based on RMSD of sidechains of residues for binding site; Retained top 5 ranked poses using 2FWZ; Ranked by top ranked pose score obtained using 2FWZ. / 0.22 / 0.057 / 0.3 / 0.08
11 / 564f8205a4c32 / Michel, Julien / Glide XP-BEDAM / 180 ligands divided into 2 groups; Ranked by merging BEDAM scores for non-aminopyrimidines; Adjusted XP scores for aminopyrimidines. / 0.21 / 0.053 / 0.29 / 0.08
12 / 564fa0ae41249 / Koes, David / qsar.ecfp6 / An ElasticNet linear model was trained on ECFP6-based fingerprint descriptors using data from ChEMBL, where the fingerprint type is varied appropriately.DOI: 10.1007/s10822-016-9960-x / 0.2 / 0.051 / 0.29 / 0.07
13 / 564f8205a4c32 / Deng, Nanjie / Glide SP-BEDAM / Generated 311 protonation states; Docking performed using the open/close conformations in the presence of 2 cavity waters; Ranked by lowest calculated binding free energy; Adjusted BEDAM scores using an empirical rule.DOI: 10.1007/s10822-016-9952-x / 0.19 / 0.053 / 0.26 / 0.08
14 / 564f8205a4c32 / Deng, Nanjie / Glide SP / Same as #13; Ranked by GLIDE SP scores.DOI: 10.1007/s10822-016-9952-x / 0.19 / 0.056 / 0.26 / 0.08
15 / 564e510834315 / Michel, Julien / rDOCK-c / Assign crystal structures to 3 clusters based on RMSD of sidechains of residues for binding site; Ranked by top ranked pose score obtained using 2CCU. / 0.19 / 0.05 / 0.29 / 0.07
16 / 564fb9530127a / Iorga, Bogdan / Cactvs_RMSD-Gold_Goldscore / VS using Glide SP; Superimposed, calculated common center of mass, virtually screened 180 ligands; Defined 3 reference ligands; Calculated RMSD, chose best RMSD and ranked by Goldscore. See: JCAMD 2016 D3R-GC Special Issue / 0.18 / 0.05 / 0.26 / 0.07
17 / 565558a40ab5f / Prabhu, Ganesh / Vina-a / Energy minimized binding site residues in the presence of HSP_90-5; Most stable conformation for each ligand used as a starting point for docking. / 0.17 / 0.046 / 0.27 / 0.07
18 / 564f4828c9e56 / Piotto, Stefano / Vina-Yada / YadaRank depends on energy as in AutodockVina as well as the proximity to well conserved residues. See: JCAMD 2016 D3R-GC Special Issue / 0.17 / 0.056 / 0.25 / 0.08
19 / 564ad0860684b / Chung, Hwan Won / VnsDock / No details. / 0.17 / 0.048 / 0.26 / 0.07
20 / 564f443622298 / Santos Martins, Diogo / Vina (3) / 1UYG and 1YC3 were used and the lowest energy pose (from Vina) selected, after applying a penalty of +0.95 to poses from 1UYG. DOI: 10.1007/s10822-016-9943-y / 0.17
(0.13 – 0.17) / 0.05 / 0.24 / 0.07
21 / 564fc1f61d439 / Rauscher, Fabian / Vina-b / GAFF force field parametersassignedto the protein and all ligands; Ligand conformational libraries generated using Omega; most stable conformation docked with Vina. / 0.17 / 0.046 / 0.27 / 0.07
22 / 564f4a42d3ee0 / Michel, Julien / Vina-c / Assign crystal structures to 3 clusters based on RMSD of sidechains of residues for binding site; Best vina-generated poses for 2CCU are present. / 0.17 / 0.055 / 0.25 / 0.08
23 / 564ca5a11a3d7 / Cho, Art / Qsite-Prime / No details. / 0.17 / 0.051 / 0.24 / 0.07
24 / 564faadc08579 / Ye, Zhaofeng / Vina-Smina-Trained_SF (4) / Structures from PDB were used to train a new scoring function; test compounds were scored by the scoring function generated and ranked by their smina scores. DOI: 10.1007/s10822-016-9941-0 / 0.16
(0.00 – 0.16) / 0.054 / 0.25 / 0.08
25 / 564e330569871 / Moro, Stefano / DockBench-Glide SP / Benchmarked 117 docking protocols; Selected pose with best (lower) pseudo-energy. See: JCAMD 2016 D3R-GC Special Issue / 0.16 / 0.054 / 0.23 / 0.08
26 / 564fa0ae41249 / Koes, David / qsar.rdkit / An ElasticNet linear model was trained on RDKit-based fingerprint descriptors using data from ChEMBL, where the fingerprint type is varied appropriately.. DOI: 10.1007/s10822-016-9960-x / 0.15 / 0.054 / 0.22 / 0.08
27 / 564ec3204a49e / Ding, Feng / MedusaDock / Used all 4 backbone conformation; grouped similar poses by clustering analysis and ranked by “free-energy” based estimation; Included internal energy changes of P & L in the estimation of binding free energy / 0.15 / 0.051 / 0.21 / 0.07
28 / 564d175b04f55 / Kumar, Ashutosh / PoPSS-Fred_CG4 score / Rescored poses using ChemGauss4; Receptors for Fred rescoring were prepared using corresponding ligand. / 0.14 / 0.051 / 0.2 / 0.07
29 / 564ec888d4faf / Park, Hwangseo / AutoDock4-VnsDock / No details. / 0.14 / 0.051 / 0.21 / 0.07
30 / 564fbfb26cb6b / Fu, Darwin / RosettaLigand / Energetically minimized starting structure without altering protein backbone; 500 models made for each PL pair; Ranked by best score observed across all models of a ligand / 0.14 / 0.053 / 0.2 / 0.07
31 / 564e50ff31d3e / Michel, Julien / rDOCK-e / Assign crystallographic structures on PDB to 3 clusters according to RMSD of sidechains residues (binding site); Added H; Visually inspected 15 highest ranked and picked 5; Ranked top poses using 4L94. / 0.14 / 0.051 / 0.2 / 0.07
32 / 564ee243b0bf4 / Anonymous / Surflex / 0.13 / 0.052 / 0.19 / 0.07
33 / 564d175b04f55 / Kumar, Ashutosh / PoPSS-Fred_CG score-ROCS_TC score / FRED Chemgauss4 score and ROCS TanimotoCombo score fused using “sum score” data fusion approach. / 0.13 / 0.051 / 0.19 / 0.07
34 / 564e3304a7724 / Moro, Stefano / DockBench-MOE_GBVI_WSA / Prepared 180 compounds and clustered according to Tanimoto similarity; Identified 4 cluster corresponding to 4 PL c omplexes; Re-scored all selected poses using GBVI/WSWA dG. See: JCAMD 2016 D3R-GC Special Issue / 0.11 / 0.052 / 0.16 / 0.08
35 / 564f443622298 / Santos Martins, Diogo / Vina-RF-Score (6) / 1UYG and 1YC3 were used and the lowest energy pose (from Vina) selected, after aplying a penalty of +0.95 to poses from 1UYG. The top pose was rescored using RF-Score and if docked in 1UYG a penalty of -0.5 to RF-Score applied. DOI: 10.1007/s10822-016-9943-y / 0.07
(0.05 – 0.07) / 0.053 / 0.09 / 0.08
36 / 564fa0ae41249 / Koes, David / qsar.smarts / An ElasticNet linear model was trained on SMARTS-based fingerprint descriptors using data from ChEMBL, where the fingerprint type is varied appropriately.
DOI: 10.1007/s10822-016-9960-x / 0.06 / 0.055 / 0.1 / 0.08
37 / 564d175b04f55 / Kumar, Ashutosh / ROCS_TC score / Scores ranked by TanimotoCombo similarity coefficient. / 0.05 / 0.056 / 0.08 / 0.08
38 / 564f7e008bfff / Gheyouche, Ennys / Vina-Qsar-b / Variant of #38. / 0.01 / 0.05 / 0.01 / 0.07
39 / 564f71dc6cc03 / Reau, Manon / Vina-Qsar-a / Docked on 2XDX & 4YKR (250 poses per ligand) & kept best energy of 1st cluster; Docked on 180 molecules and on 2XDX & 4YKR (8 poses); Ranked molecules using built linear mode. / 0.01 / 0.05 / 0.01 / 0.07
40 / 564f7cd8ed8c0 / Chan Yao Chong, Maud / Vina-Qsar-c / Variant of #38. / 0.01 / 0.049 / 0.01 / 0.07
41 / 564fb9530127a / Iorga, Bogdan / Cactvs_RMSD-Glide SP / Superimposed, calculated common center of mass, virtually screened 180 ligands; Defined 3 reference ligands; Calculated RMSD and ranked by Glide SP. See: JCAMD 2016 D3R-GC Special Issue / -0.03 / 0.055 / -0.03 / 0.08
Table S6. Methods used for Stage 2 HSP90 ligandranking.The Method ID corresponds to the numbers in the in-text figures, the Submission ID is a unique ID given to each prediction upon submission, the Participant’s Name is the last and first name of the participant, the Method Name is the informal name given to the procedure used while Method Descriptions are condensed from submitted protocol descriptions. For submission files with multiple variants of the same method, only the highest score from each such submission is reported; the names of the methods for such submissions have number of prediction sets in the method name (parenthesis) and the scores range in the scores (parenthesis).