This tendency is consistent with the preferential uses of statistical potentials for threading or folding prediction at low sequence identity and of physics based force fields for the refinement of models close to native conformations. This dichotomy suggests that model selection could be improved if we could predict which criterion to use, either selleck chem Cabozantinib MM GBSA for models closer than 1. 5 to native structure or SC3 for more distant models. How ever, such a close distant model classifier would need to be quite accurate since misclassifications would rapidly cancel the small gain obtained using MM GBSA for close models. Model database and server The 1621 known knottin sequences were extracted from the latest release of the KNOTTIN database.
A struc tural model of each knottin sequence was built using the optimized procedure detailed above 20 templates were selected according the TMS criterion and without restric tion on the query versus template sequence identity. These templates were multiply aligned with the query sequence using the TMA procedure. Then, using from 1 to 20 aligned templates, 5 structural models of the query were generated at each Modeller run after imposing appropriate constraints on the knotted disulfide bridges and the 80% conserved hydrogen bonds. The 20 Modeller runs resulted in 100 structural models per query which were sorted according to the SC3 criterion. Finally, the energy of the best model was minimized using the sander program of the Amber package. Restraints were applied on the backbone atoms to avoid large deviations from the initial model and the GBSA implicit solvation scheme was used.
Further difficulties arise when attempting to automatically model large data sets. Since several knot tins are macrocyclic, i. e. the N and C termini are con nected through a regular peptide bond, potentially cyclic knottins were tentatively modeled as such according to the annotation available in the KNOT TIN database. In the latter database, the cyclic feature was assessed by manually analyzing the N and C termini for the presence of a cyclization site. Moreover, a large number of knottins display additional disulfide bridges that supplement the 3 disulfides forming the cystine knot. These additional bridges were only imposed in the models when there was no ambiguity regarding cysteine connectivity. In any case, when residues at standard posi tions 82 and 98 were cysteines, a disulfide bridge was always imposed whatever the total number of cysteines, since this bridge has been frequently observed in experi mental Anacetrapib structures. Finally, except for knot tins with known 3D structure, the resulting knottin structural models are now available from the Sequence section of the KNOTTIN database server at URL.