PROTEIN LIGAND USEFUL SIMULATION
The need for structural information for fragment-based design of bio-active compounds
The growing popularity of Fragment-Based Drug Design (FBDD) shows that this methodology is more and more recognized as a tangible alternative to high throughput screening, and a successful method for hit identification and for lead conception. The methodology is also reported as a novel powerful method for Chemical Biology.
The FBDD approach consists of identifying low-molecular weight compounds (fragments) that weakly bind to the target protein. These very simple molecules with few chemical functional groups usually display low affinity for the target (high μM to low mM), compared to the bigger, more complex molecules used in classical HTS campaigns. Optimization of the fragment hits, usually by addition of new chemical functions or by linking of two fragment hits binding in adjacent pockets, is strongly driven by structural information.
X-Ray crystallography is the key technique for this structure-based drug design process, but its applicability is strongly case dependent, due to the low affinity of the fragments. Docking calculations can also be done but they still suffer from the fact that the scoring functions used to rank the docked ligand positions are not optimized enough for interactions involving small and weak ligands such as fragments. Therefore, new methods including experimental data are required to allow a fast evaluation of the binding site and binding mode of the fragments once they have been detected as ligands in screening experiments. Nuclear Magnetic Resonance (NMR) is one of the most powerful techniques to screen and identify hit fragments. The method also enables the fragment binding site identification, thanks to qualitative interpretation of the Chemical Shift Perturbation (CSP) observed in the protein NMR spectra upon addition of the fragment.Here, we use proton CSP quantitatively to obtain 3D structures of protein-ligand complexes by the combination of docking and CSP data. We use the 3D structure of the free protein or possibly a 3D structure of the protein bound to other ligands.
We have developed a program CSPSim for the back-calculation of CSP from virtual ligand positions in the protein structure generated by docking, the comparison with experimental CSP values, and the ranking of the ligand positions according to the agreement between experimental and back-calculated data.
We have also modified the docking program PLANTS to allow guided-docking using CSP data. Here the comparison of the CSPsim and CSPexp values is directly included into the docking process as a CSP factor is part of the scoring function.
We have shown that protons do not all perform similarly. In particular Ha protons lead to better results compared to amide protons.
We have also shown that the comparison of the CSP profiles for the docking ensemble (> 200 poses) with the experimental CSP profil is crucial to understand whether:
- The protein undergoes conformational change upon ligand binding.
- One particular residue should be removed from the calculation due to large experimental CSP value not explained by any or the position, and likely related to a change in the hydrogen bond network upon ligand binding
We show here, using the protein-ligand complex 1WUG, how the CSP-guided docking results are modified when the residue Y809 is removed from the CSP back-calculation (pink structure) or not (yellow structure) during the docking. The experimental structure is shown in green for comparison.
This ANR project has generated new tools for the analysis of protein-ligand complexes, providing new insights into the complexity and versatility of low-affinity protein-ligand interactions, and will have an impact both in the field of Chemical Biology and Drug Design.