Innovative Approaches to Computational Drug Discovery
EPFL on iPhone
- Marco De Vivo (Italian Institute of Technology, Genova, Italy)
- Andrea Cavalli (Italian Institute of Technology, Genova and University of Bologna, Italy)
- Matteo Dal Peraro (Swiss Federal Institute of Technology Lausanne (EPFL) , Switzerland)
Computational methods are nowadays essential to all aspects of designing and optimizing a new drug. Computational tools are routinely exploited in the target discovery, hit identification, hit-to-lead, and lead optimization phases of a drug discovery project. The role of computation has recently been strengthened by the rapid development of faster architectures and better algorithms, which have created the real possibility of applying high-level computations in a time-affordable manner. For instance, enhanced sampling calculations can account for the full flexibility of the target-ligand complex. Previously a pure computational exercise, they now can play a key role in drug design, providing valuable information on ligand-binding affinity and helping researchers to discern active compounds from inactive ones .
Computational drug design approaches can be roughly divided into ligand-based and structure-based methods. The available physicochemical information dictates what strategy is more likely to be applied. Where structural data of the target protein exist, structure-based drug design [SBDD] is by far the most applied strategy. However, when this structural information is missing or not fully reliable [e.g., homology models of the target based on poor sequence identity], ligand-based approaches to drug design are usually applied. These latter approaches are also usefully deployed when biological information is available for a large number of compounds. For instance, toxicity data for hundreds of thousands of molecules can be used to extract mathematical models for early toxicity prediction of novel drug candidates. Simulations of complex systems at the cellular and organ level, synthesis route prediction, and prodrugs development are also important fields in which computation can play a key role. Clearly, a computational chemist wishing to succeed in drug design must be familiar with the full variety of computational approaches available .
SBDD is one area of interest for the proposed workshop, with focus on the major challenges and advances in this field. Several methods can be considered as part of the computational armamentarium for SBDD, spanning from docking calculations to classical molecular dynamics [MD]. However, each method has its limitations. Docking is quite good at correctly posing compounds into the cavity, but scoring remains a major issue. Is ranking simpler than scoring? If so, can we use it? For instance, a recent application of steering MD has been able to rank compounds, and to discern active from inactive ones . This seems a promising strategy for hit identification. Nevertheless, a major challenge of computation in predicting active compounds is to find the right balance between accuracy and speed. Nowadays, CPU-intensive calculations, such as free energy perturbation (FEP), can be used to accurately estimate the binding free energy . However, the sampling and convergence of the results are not always complete, and it is still difficult to handle large structural differences between ligands. Furthermore, depending on the stage of the project, other issues might arise, which are not related to the binding of the drug to the target. For instance, the physicochemical properties of active compounds dictate their pharmacokinetics (PK) profile. Adsorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) are key parameters to be tuned in order to produce a drug candidate with a drug-like PK profile. Can we include ADMET prediction in an early stage, thus preventing the possibility that a novel drug candidate will fail in the later costly clinical phases? In silico ADMET prediction remains a major challenge in drug discovery.
Finally, quantum mechanics (QM) has recently shown itself ready to make an impact on SBDD [5-8]. From QM-based scoring function, to QM studies of enzymatic mechanisms for transition state analogues design, QM methods now have real potential for drug discovery. From statistical mechanics to quantum mechanics, physics-based methods can play a major role in improving the performance and predictivity of computational drug discovery methods . These major aspects of computation will be discussed during the workshop, which will benefit from an audience that includes both academic and industrial scientists.
 W.L. Jorgensen,‘Drug discovery: Pulled from a protein's embrace’, Nature 466, 42-43 (2010).
 W.L. Jorgensen, ‘The many roles of computation in drug discovery’, Science 303, 1813-1818
 F. Colizzi, R. Perozzo, L. Scapozza, M. Recanatini, A. Cavalli, ‘Single-molecule pulling
simulations can discern active from inactive enzyme inhibitors’, J. Am. Chem. Soc. 132, 7361- 7371 (2010).
 W.L. Jorgensen, ’Efficient drug lead discovery and optimization’, Acc. Chem. Res. 42, 724-
 T. Zhou, D. Huang, A. Caflisch, ‘Quantum mechanical methods for drug design’,
Curr. Top. Med. Chem. 10, 33-45 (2010).
 K. Raha, M.B. Peters, B. Wang, N. Yu, A.M. WollaCott, L.M. Westerhoff, K.M. Merz, ‘The
role of quantum mechanics in structure-based drug design’, Drug Discovery Today 12, 725-
 M. Dal Peraro, P. Ruggerone, S. Raugei, F.L. Gervasio, P. Carloni, ‘Investigating biological
systems using first principles Car-Parrinello molecular dynamics simulations’, Curr. Opin. Struct. Biol. 17, 149-156 (2007).
 A. Cavalli, P. Carloni, M. Recanatini, ‘Target-related applications of first principles quantum
chemical methods in drug design’, Chem. Rev. 106, 3497-3519 (2006).
 M. De Vivo, M. Dal Peraro, M.L. Klein,‘Phosphodiester cleavage in ribonuclease H occurs via
an associative two-metal-aided catalytic mechanism’, J. Am. Chem. Soc. 130, 10955-10962
 Y. Deng, B. Roux, ‘Computations of standard binding free energies with molecular
dynamics simulations’, J. Phys. Chem. B. 113, 2234-2246 (2009).