Chemoinformatics approaches for new drugs discovery

Chemoinformatics uses computational methods and technologies to solve chemical problems. It works on molecular structures, their representations, properties and related data. The first and most important phase in this field is the translation of interconnected atomic systems into in-silico models, ensuring complete and correct chemical information transfer. In the last 20 years the chemical databases evolved from the state of molecular repositories to research tools for new drugs identification, while the modern high-throughput technologies allow for continuous chemical libraries size increase as highlighted by publicly available repository like PubChem [http://pubchem.ncbi.nlm.nih.gov/], ZINC [http://zinc.docking.org/], ChemSpider[http://www.chemspider. \ncom/]. Chemical libraries fundamental requirements are molecular uniqueness, absence of ambiguity, chemical correctness (related to atoms, bonds, chemical orthography), standardized storage and registration formats. The aim of this work is the development of chemoinformatics tools and data for drug discovery process. The first part of the research project was focused on accessible commercial chemical space analysis; looking for molecular redundancy and in-silico models correctness in order to identify a unique and univocal molecular descriptor for chemical libraries indexing. This allows for the 0%-redundancy achievement on a 42 millions compounds library. The protocol was implemented as MMsDusty, a web based tool for molecular databases cleaning. The major protocol developed is MMsINC, a chemoinformatics platform based on a starting number of 4 millions non-redundant high-quality annotated and biomedically relevant chemical structures; the library is now being expanded up to 460 millions compounds. MMsINC is able to perform various types of queries, like substructure or similarity search and descriptors filtering. MMsINC is interfaced with PDB(Protein Data Bank)[http://www.rcsb.org/pdb/home/home.do] and related to approved drugs. The second developed protocol is called pepMMsMIMIC, a peptidomimetic screening tool based on multiconformational chemical libraries; the screening process uses pharmacophoric fingerprints similarity to identify small molecules able to geometrically and chemically mimic endogenous peptides or proteins. The last part of this project lead to the implementation of an optimized and exhaustive conformational space analysis protocol for small molecules libraries; this is crucial for high quality 3D molecular models prediction as requested in chemoinformatics applications. The torsional exploration was optimized in the range of most frequent dihedral angles seen in X-ray solved small molecules structures of CSD(Cambridge Structural Database); by appling this on a 89 millions structures library was generated a library of 2.6 x 10 exp 7 high quality conformers. Tools, protocols and platforms developed in this work allow for chemoinformatics analysis and screening on large size chemical libraries achieving high quality, correct and unique chemical data and in-silico models