Figure 4
Use of PRODRG-generated topologies. (a) GlcNAc5 in ChiB (van Aalten et al., 2001). Left, stereo diagram of the ligand molecule before (cyan) and after (green) refinement with a PRODRG-generated topology. The surrounding protein is shown as a semitransparent cartoon. Right, text drawing used to generate the topology. (b) As (a) for Triton X-100 in SCP-2L (Haapalainen et al., 2001). (c) Ligand from a high-resolution structure (cyan molecule) of human neutrophil collagenase (Gavuzzo et al., 2000) re-refined at lower resolution with topologies generated either with PRODRG (green molecule) or with LIBCHECK (orange molecule). Again, the protein is shown as a semitransparent cartoon. To the right, the chemical structure of the ligand [2-(biphenyl-4-sulfonyl)-1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid] is given. (d) As (c) for (3-amino-2,5-dioxo-1-pyrrolidinyl)-acetic acid in Cryphonectria parasitica endothiapepsin (Erskine et al., 2003). (e, f) Effect of poor input geometries on the quality of generated topologies as indicated by the r.m.s.d. between small-molecule coordinates from the `ideal' starting structure and the same structure after refinement at lower resolution. The refinement of HGPRT as described in Table 3 is repeated with topologies generated from coordinates perturbed by a given random shift (filled squares). In addition, the corresponding refinement results using topologies produced in a coordinate-independent manner are given (empty diamonds). For PRODRG this means topologies were generated from TXT-mode drawings; for XPLO2D the topologies available from HIC-Up were used and for LIBCHECK the ligands were drawn in SKETCHER. Weights are kept at the values given in Table 3. (e) shows the results for refinement with CNS and (f) for REFMAC5. |