|
|
|
|
| Machine Learning & Graph Mining |
|
eCheminfo InterAction Meeting Session, Basel, 10 November 2005 (proceedings available through website) Applications of Machine Learning & Graph Mining in Drug Discovery chaired by Stefan Kramer (Technische Universitaet Muenchen) eCheminfo 2005 InterAction Meeting, 9-10 November 2005, Basel, Switzerland
Presenters & Discussion Leaders:
Lazy-Structure-Activity-Relationships (lazar) for the in-silico Prediction of Chemical Carcinogenicity, Christoph Helma (University of Freiburg)
Graph-Theoretic Procedures for Searching Databases of Molecular Structures, Peter Willett (University of Sheffield)
Finding Discriminative Substructures Using Elaborate Chemical Representation, Joost N. Kok (Leiden University)
Alignment-free Potential Pharmacophore-Point Descriptors for "Informed" Similarity Searching, Gisbert Schneider (Johann Wolfgang Goethe-University)
Mining of Molecular Fragments with Wildcards, Michael Berthold (University of Konstanz)
Comparing Molecule Similarity Codings - Vectorial versus Structural Coding for ADME/Tox Prediction, Joerg Kurt Wegner (Tibotec BVBA)
The goal of this group and meeting session is to bring together researchers from graph mining and computational chemistry/cheminformatics/drug design. Graph mining experts will report on the state of the art and the possibilities offered by new techniques developed recently in this rapidly evolving field. Experts from cheminformatics and related areas will give an account of the challenges and pitfalls for graph mining techniques applied to chemical data. The technical part will cover issues such as: vertex and edge labels other than nominal, aggregates, 2D vs. 3D, feature selection, search techniques and pruning, overfitting avoidance, chemical similarity, lazy and instance-based learning, combining structured coding approaches with 3D information, including descriptions of pharmacophores and bioisosters based on field-based approaches, combining different coding approaches of ligands and protein targets, like consensus scoring, ensemble learning and multi-objective learning for ADME/Tox prediction, high-throughput docking studies and protein shape. The application areas will cover drug design and pharmocology, predictive toxicology and toxicogenomics, etc. The overall outcome of the program should be a set of matches between techniques and potential applications: Which new techniques could be useful where? Which new graph mining techniques can be readily applied to solve problems in cheminformatics? Which issues should be addressed by graph mining research? What are the challenges for graph mining approaches to be useful in real-world applications in this area?
Posters All registrants and members are eligible to submit a Poster. Attendees may view and discuss the Posters and leave messages for the authors on the Web site. Poster Abstracts (of ca. 300 words) with Title, Institution, Authors and Contact Information should be submitted for consideration to echeminfo at douglasconnect.com Posters can be presented as HTML, pdf, Powerpoint or Word documents.
|
|
|
|
|
|
|
