, Drug Design 2012, 1:2 http://dx.doi.org/10.4172/2169-0138.S1.002

International Conference and Exhibition on Computer Aided Drug Design & QSAR October 29-31, 2012 DoubleTree by Hilton Chicago-North Shore, USA

Recent technological advancements in ligand-based and structure-based modeling: BCL: Cheminfo and rosetta ligand Edward W. Lowe Jr. Center for Structural , USA

oth ligand-based and structure-based computational methods are invaluable tools in computer aided drug discovery. BThis presentation will introduce technological advancements in both disciplines. First, we introduce BCL: ChemInfo, a comprehensive machine learning-based quantitative structure activity relationship (QSAR) modeling framework featuring novel molecular descriptors, diverse automated feature selection, GPU acceleration, and consensus model analysis. Public availability of high-throughput screening (HTS) data is rapidly increasing highlighting the need for ligand-based computational methods, such as BCL: ChemInfo, to accelerate probe development and drug discovery while reducing costs. The framework was benchmarked on publically available HTS data (PubChem) for nine targets, selected as representatives for the major protein families most commonly targeted by therapeutics, to investigate the influence of size and composition of training data, the choice of objective function, effectiveness of several feature selection techniques, as well as the predictive power of consensus predictors using different machine learning techniques. Second, recently advancements in the small molecule docking using RosettaLigand will be presented including simultaneous docking of explicit interface water molecules, and small molecule docking into comparative models. A study performed on protein-centric water docking shows an improvement in ligand placement at a ratio of 9:1 while ligand-centric water docking allows for the recovery of up to 56% of failed docking studies using 341 structures from the CSAR benchmark of diverse protein-ligand complexes. A study on docking into comparative models found RosettaLigand was successful in recovering a native-like binding mode among the top ten scoring binding modes for 21 of 30 cases while template selection based on ligand occupancy rather than template-target identity was discovered to increase success.

Biography Edward W. Lowe Jr. has completed his Ph.D at the age of 26 years from the University of South Florida and postdoctoral studies from Vanderbilt University Department of . He was an American Heart Association pre-doctoral Fellow, a postdoctoral trainee on the NIH roadmap Integrative Training in Therapeutic Discovery fellowship, and more recently, an NSF Cyber-Infrastructure Transformative Computational Science Fellow. He is now Research Assistant Professor of Chemistry at Vanderbilt University leading computational methods development in and computational drug discovery efforts in the laboratory of Jens Meiler.

[email protected]

Drug Design 2012 CADD-2012 Volume 1 Issue 2 ISSN: 2169-0138, an open access journal October 29-31, 2012 Page 27