Introduction
The NFCR Center for Computational Drug Design is a “virtual center” founded by the the Bethesda, Maryland based National Foundation for Cancer Research (NFCR). (http://www.nfcr.org/) It is directed by Professor Graham Richards (Oxford University, UK) and involves satellite units in Spain, Portugal, England and Italy linked by the internet. (http://www.chem.ox.ac.uk/ccdd/ccdd.html). These groups share a common goal: the design of small molecules as potential drugs which bind to and inhibit the action of proteins involved in cancer or have nucleic acids as the target. The internet and world wide web are used to make the researcher’s interaction instant, therefore each unit has access to a range of expertise and facilities that could not be assembled in one place. All the groups are taking part in the flagship project of the center. This is the widely publicised screensaver project to screen 3.5 billion small molecules as potential inhibitors of 16 proteins which are current targets of the pharmaceutical industry (http://www.ud.com).
The development of a new drug is a multi-step process. First, an enzyme or receptor involved in a disease needs to be identified. Second, the structure of the enzyme or receptor has to be established. Finally, a molecule which will bind tightly to the enzyme or receptor must be designed and synthesised. The explosive growth of information provided by the cloning, sequencing and expression of a variety of receptors, on the experimental side, and the increased availability of fast computers and improvements in the computational methodologies, on the theoretical side, allow the theoretical investigation of structural and functional features of large proteins, bringing researchers one step closer to the understanding of the mechanisms triggered by molecular recognition and leading to signal transduction. This knowledge can be exploit to optimise the binding energy of the small molecule to its receptor and hence reduce the drug dosage required, or to design into the drug a high level of selectivity towards its target receptor binding site or enzymatic active site.
The research of the Italian satellite unit focus on the definition and determination of theoretical molecular descriptors for small biologically active molecules and their comparison on the ground of predictive ability and physical interpretability. The problem of relating structure to activity in order to achieve quantitative structure-activity relationship (QSAR) models easy interpretable in terms of reactivity concepts and intermolecular interactions is of fundamental importance in rational drug design. To this respect, computational simulations of ligand-receptor or protein-protein interactions allow the definition of intermolecular interaction descriptors, which possess high information content and might help, through the formulation of statistic models, the rationalisation of the observed biological responses (affinity, selectivity, efficacy, etc.). Therefore, the prediction of biological properties before experimental testing, and the suggestion of further lines of experimental inquiry is possible. In fact, the theoretical studies are designed to complement experimentation in providing mechanistic understanding at the molecular level of detail, and to guide pointed experimental exploration of cellular processes and functions in numerous collaborative studies.

Methods and Approaches
The approaches we use include theoretical determinations of molecular structure and properties, and computational simulations of molecular mechanisms and processes. The computational procedures we elaborate are continuously being refined and tested with the experimental results obtained for the biomolecular systems under study. They are based on methods of quantum and statistical and molecular mechanics, and implemented in algorithms often developed by other research units of the network running on supercomputers and graphic workstations. Some of the topics currently in hand are described below.

Quantitative Structure-Activity Relationships(QSAR).
Several problems are connected with the attempt to rationalise in quantitative terms the pharmacological data which concern receptors. Protein biochemistry approaches combined with DNA cloning and sequencing techniques have demonstrated that receptors constitute families of proteins built on a common structural scheme, despite the wide molecular diversity of their ligands. Thus, subtle molecular determinants might be responsible for the selectivity towards different receptor subtypes, and apparently small chemical modifications of a ligand can change its pharmacological phenotype dramatically. Moreover, although cells expressing cloned receptors constitute useful tools for drug screening and facilitate the establishment of precise structure-affinity relationships, the ultimate goal is to generalise the information obtained on a pure population of receptors to the native receptors in order to design tissue-selective drugs. The process of distinguishing the few functionalities that contribute to binding from those that are superfluous and are responsible for side effects, depends primarily on the availability of descriptors able to capture the strict ligand-receptor complementarity criteria, which determine the biological properties of interest. The aim is to define theoretical molecular descriptors which give powerful predictive and interpretative QSAR models to be used for drug design as well as theoretical (in-silico) screening of in-vitro and in-vivo pharmacological data.

Protein assemblies
The nucleation of multimeric complexes seems to be a common biological strategy for cytokines, which share structural or functional similarities with IL-6, utilising the common signalling chain gp130. Deciphering the pathway of receptor complex formation and identifying the patches of amino acid residues on the ligands and receptors which constitute the driving force of this process is of fundamental importance both for understanding the mechanisms responsible for the activation of intracellular processes, and for the generation of small ligand mimics with specific biological properties. In fact, the main aim of this project is to obtain peptides and peptido mimetics as potential drugs to be employed in many pathologies. At the moment, biological screening of a number of peptides designed on the basis of the computational simulations carried out on the IL-6 multimeric complex are under biological investigation on IL-6 dependent plasmocytoma and breast cancer cell lines. (Fig. 1)

Substrate specificity
Many known carcinogens are specific for cytochromes of the P450 I family. In particular, polyaromatic and heteroaromatic hydrocarbons and their amino derivatives are metabolised by the P450 I family of enzymes yielding to reactive intermediates which can interact with DNA, forming covalent adducts that bring miscoding, mutagenesis and carcinogenesis. Quantitative theoretical investigations of substrate specificity towards cythocromes P450 are mainly carried out by means of substrate-based approaches. However, intermolecular interaction descriptors computed on the cytochrome/substrate complexes can be used to gain insights into the molecular determinants for substrate specificity in connection with the structural and topological differentiation of cytochromes. The main goal is the elaboration of quantitative models for predicting specificity and carcinogenesis of new compounds.

Ligand-receptor interactions
Recent findings showed an increase in specific binding of peripheral benzodiazepine receptor (PBR) ligands in certain human brain tumors. This prompted us to undertake an iteractive synthetic-computational approach to the study of the binding site of PBR ligands in order to obtain putative radioligands and specific carriers for antitumor drugs. (Fig. 2) The combined use of different computational approaches carried out both on the isolated ligands and on the ligand-receptor complexes provided a quantitative rationalisation of the structural characteristics of PBR antagonists and a detailed picture of the interacting molecule complementarity. Speculations on the mechanistic role of the main structural components are, therefore, possible.

Conclusions
The rapid progresses in the techniques of molecular biology and the impressive increase of crystal and solution NMR structures of biological macromolecules made available in these last years have increased the productivity of data generation. Turning the huge amount of data accumulated into knowledge contributing to drug discovery is mandatory. The advanced molecular simulation methodologies and expertises available at the Virtual Centre for Computational Drug Design are utilised for the development and validation of sophisticated computational protocols for the rationalisation of the experimental data and comprehension of biomolecular functions. The Italian research unit contribution constists mainly in the elaboration of quantitative protocols for the theoretical screening of biological properties expressed by small ligands in the interaction with their enzyme/receptors, as well as by interacting macromolecules. Owing to their interpretative and predictive power these protocols, used in close connection with experimental investigations, gain an important role in the study of the biochemical processes at molecular level and in rational drug design.

Maria Cristina Menziani
Dipartimento di Chimica
Università di Modena e Reggio Emilia