Prediction of biochemical properties

HexX (http //www.biosolveit.de/flexx) is a fast, robust, and highly configurable (FlexX-able) computer program for predicting protein-ligand interactions. Its main application is 

AutoDock (http //autodock.scripps.edu/) which is designed to predict how small molecules, such as substrates or drag candidates, bind to a receptor of known 3D structure. AutoDock has appUcations in X-ray crystafiognphy, lead optimization, structure-based design, combinatorial chemistry, protein-protein docking and chemical mechanism studies 

GRID (http //www.moldiscovery.com/soft grid.php) is a computational procedure for determining energetically favorable binding sites on molecules of known structure. It may be used to study individual molecules such as drags, molecular arrays such as membranes or crystals, and macromolecules such as proteins, nucleic acids, glycoproteins or polysaccharides. Several different molecules can be processed one after the other. 

Lastly in this section, a new improvement to the JME molecular editor (see above, http //www.molinspiration. com/cgi-bin/properties) permits prediction of bioactivity. 

There is a lot of information here, in a ready-to-use format. For example, blood-brain barrier penetration data is available on a training set of 57 compounds and a data set of 13 more long-term animal carcinogenicity results are available for over 1,4(X) compounds, drawn from the Carcinogenic Potency Database (CPDB), an initiative of the Lawrence 

---

The kernel function measures the similarity between pairs of patterns, typically as a dot product between munerical vectors. The usual numerical encoding for protein sequences is based on a 20-digit vector that encodes (binary) the presence/absence of a certain amino acid in a position. To explore new ways of encoding the structural information from biosequences, various kernels have been proposed for the prediction of biochemical properties directly from a given sequence. 

---