journal menu
![[Figure 1]](ud5027fig1mag.jpg)
|
|
Figure 1
Frag4Lead workflow for fragment growing. The starting point of the workflow is a crystallographically detected fragment hit. It provides two types of information. The first is the identity, i.e. the chemical structure, of the fragment hit, based on which potential follow-up candidates are retrieved from the commercial catalog of MolPort and 3D conformers generated by OMEGA. The second is the 3D information of the binding pose of the fragment hit inside the binding pocket. The binding pocket is then prepared as a docking receptor via the LeadIT software (see Section 2.3 ![[link]](../../../../../../logos/arrows/d_arr.gif) for details). Template-guided docking is then employed via a customized script using FlexX (Rarey et al., 1996) using the crystallographic binding pose of the fragment as a starting point. Specifically, the FlexX algorithm cleaves each analog into internally rigid fragments (referred to as `FlexX fragments'). The FlexX fragment most similar to the starting fragment is then superimposed on the latter. Finally, each analog is incrementally reassembled by flexibly attaching its constituent FlexX fragments to the superimposed base fragment and the binding site is explored by FlexX docking. A maximum of 100 docking poses for each analog are generated and only the 1000 highest-scoring analogs are considered. In rare cases this process needs manual intervention, for example pruning of the docking template. The next vital step in the Frag4Lead workflow is the processing of the docking results. The docking poses generated by FlexX are rescored by the DrugScoreX per-contact score (see Section 2.4 for details). Next, redundant docking poses that are very similar to a better scored retained pose and would otherwise complicate the assessment of relevant poses are removed. To this end, the following procedure is applied to the docking poses of each analog. Firstly, all poses are clustered by hierarchical complete-linkage clustering with an r.m.s.d. threshold of 2.0 Å as implemented in fconv (Neudert & Klebe, 2011b). Only the three best-scoring, nonredundant and internally sorted poses are kept. This efficiently eliminates redundant poses and allows the direct comparison of unique poses of each analog. In order to present the ranked hit list for interactive evaluation in a way that is also amenable to non-expert users, a PyMOL session is created that highlights the interactions and per-atom contributions (green spheres) to the overall DrugScoreX score of the pose. Unfavorable interactions and contributions are likewise highlighted. This enables a convenient and informed selection of follow-up compounds to be acquired based on the following criteria: (i) the ability of an analog to bind in the corresponding fragment pose, (ii) the location of most of its structure in a favorable environment, indicated by high but evenly distributed per-atom contributions to the overall DrugScoreX score, (iii) the formation of additional or alternative interactions compared with the starting fragment and (iv) the adoption of a realistic conformation. The binding of acquired compounds is then investigated by X-ray crystallography. The blue mesh shows the 2mFo − DFc electron-density map for the follow-up ligand contoured at σ = 1.0. Observed binders are then further evaluated by ITC to assess their binding affinity. |
![[Figure 1]](ud5027fig1.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
Open 
access
access
