Home | Search | Help  
Home Page Università di Genova

Seminar Details

Date 28-4-2009
Time 16:00
Room/Location Al DISI, Sala conferenze 3 piano
Title Computational Challenges in Drug Discovery
Speaker Dr. Andrea Cavalli - Dr. Walter Rocchia
Affiliation IIT
Link https://www.disi.unige.it/index.php?eventsandseminars/seminars
Abstract Computational Challenges in Drug Discovery The Computational Chemistry Group of the Italian Institute of Technology will describe its research activity outlining some open issues for the scientific community in the field of Drug Discovery. Hereafter, two areas where several among them arise are briefly sketched. The quantitative structure-property relationship A quantitative structure-property relationship (QSPR) attempts to correlate an experimentally measured property to some topological and/or physicochemical descriptors derived from the molecular structure. In drug discovery, QSPRs have been widely applied to correlate biological activities to a molecule’s electronic characteristics and hydrophobicity, through the use of one-dimensional molecular descriptors and simple statistical methods, such as linear discriminant analysis. When three-dimensional features need to be taken into account, a set of conformations (one for each considered molecule) is overlaid following a binding mode hypothesis. The molecular fields surrounding each molecule are estimated on a regular lattice that encompasses the molecule. Here, since the number of descriptors is much larger than the compounds, partial least-squares (PLS) regression is usually used to find a correlation between field values and biological responses. In this context, a new generation of QSPRs is advisable, and it should make use of more robust statistical methods encompassing a panel of plausible bio-active conformations for each molecule also taking into account new possible metrics to evaluate molecular differences. Binding Free Energy Characterization A further strategic topic is the description of the binding kinetics of chemical compounds. In this context a thorough characterization of the potential of mean force (PMF), or, in other words, of the Free Energy of binding, along a suitable set of reaction coordinates is crucial. In order to obtain accurate results, a full sampling of the whole conformational space would be required, but this is not feasible with current computational facilities. This is the rationale for the so-called “enhanced sampling” techniques, which aim at sampling only the conformational states that are relevant to the binding event, through a careful and often far from obvious choice of the reaction coordinates. Available criteria, which are based on the physical comprehension of the binding dynamics, are hardly generalisable and rely too much on the physical intuition of the user. Hence, maximum likelihood techniques and Bayes’s relation based approaches have been applied to identify the most relevant descriptors of the small molecule docking process. Information driven identification of such relevant descriptors is therefore of vital importance in the drug discovery process since it will lead to more effective reaction coordinates. Along with this issue, many limitations come from the model used in describing the ligand-target interaction. The models used so far are effective but lacks of polarization effects which would be possibly addressed by an information-driven fitting procedure. The IIT Computational Chemistry Group: Our group is presently composed by Andrea Cavalli, our group leader, as well as Walter Rocchia, team leader, and several post-docs, namely Angelo Favia, Giovanni Bottegoni, Matteo Masetti and Davide Branduardi. Our group provides support to all project teams at the IIT Drug Discovery and Development platform. It applies standard ligand- and structure-based approaches to speed up the discovery of novel drug candidates. Docking and statistical analysis are usually exploited in the hit identification, hit-to-lead, and lead optimization phases of a drug discovery project. At the same time, the group works on innovative methods to enhance the predictive power of computational tools in drug discovery. Our facilities are equipped with state-of-the-art hardware, including several quad-core workstations and a server endowed with a computational power approaching 2 Tera-Flops.
Back to Seminars