Supplementary Material S1 - Towards Improved Quality of GPCR Models by Usage of Multiple Templates and Profile-Profile Comparison

This file contains: Data sets used in the study, Tables S1-S3, Figures S1-S5 and Supplemental Methods. Table S1. The templates data set used in the current study. Table S2. GPCRs released in 2012 - benchmark results of web services in GPCR structure modeling. Table S3. GPCRs released in 2012 - benchmark results of GPCRM in GPCR structure modeling depending on the templates data set. Figure S1. The reference sequence alignment of GPCRs. The alignment was generated by VMD (a MultiSeq plugin [22]) based on the structural alignment of GPCRs of known 3D structures. Positions of highly conserved residues are marked according to Ballesteros-Weinstein numbering scheme. Positions of TM helices based on rhodopsin (1GZM) are marked with grey. Figure S2. The sequence alignment used in GPCRM modeling of A2AR. A fragment which corresponds to the bulge in TMH4 is marked by a square box. The template with the bulge in the structure (2VT4 – β1AR) is aligned against the target sequence (A2AR) without any gaps in that fragment while the template without the bulge (3RZE – H1R) is aligned with a one-residue gap. Figure S3. The sequence alignment used in GPCRM modeling of κ-opioid receptor. A fragment which corresponds to the lack of bulge in TMH2 is marked in the alignment (a square box). The template without the TMH2 bulge (3ODU – CXCR4) is aligned against the target sequence (κ-opioid receptor) without any gap in that fragment, while the template with the TMH2 (3RZE – H1R) bulge is aligned with a one-residue gap. Figure S4. The model of κ-opioid receptor (PDB id: 4DJH). The model (green) was generated by GPCRM and superposed on the crystal structure (blue) and templates used in the model building: the histamine H1R (grey) and the CXCR4 receptor (pink). The bulge observed in TMH2 in H1R was removed and was not transferred to the κ-opioid model. Nevertheless, averaging of H1R and CXCR4 coordinates in TMH1 did not result in the proper kink of TMH1 proving limitations of the Modeller software. Figure S5. Models of κ-opioid receptor (4DJH) generated by currently available methods. All models are superposed on the crystal structure (blue). The bulge in TMH2 which is not present in the crystal structure is depicted. Templates used in the model building by each method are as follows: rhodopsin (ModWeb/ModBase), β1AR (GPCRDB and GPCR-Modsim), β1AR together with β2AR, A2A and rhodopsin (SSFE). Figure S6. Ligand docking to GPCRM-generated homology models versus self-docking: β2AR (A), H1R (B), CXCR4 (C) and metarhodopsin II (D). The reference crystal complexes with indicated polar contacts (yellow dashed lines) are shown in grey, while the docked ligand poses are depicted in yellow. GPCRM-generated homology models of receptors are shown in green. Left panels show the best poses obtained from docking to corresponding protein homology models. Right panels show results of self-docking to crystal structures (PDB id: 3SN6, 3RZE, 3ODU, 3PQR). Most polar contacts were preserved except for: Ser203 (A), Thr112 (B), Asp97 (C). Although Ile189 and Tyr191 in the EC2 loop are not as deep in the binding pocket as in the crystal structure of metarhodopsin II (D), retinal was positioned in the homology model with the proper orientation of the β-ionone ring (left panel) contrary to the self-docking results (right panel). (DOCX)