We use kernel functions for inferring similarity of agiven set of compounds, for classification or regression depending onapplication.  We also explore whataspects of chemical compounds can help us encode them better. We attempt toimprove sequence/graph representation at the 2D through 4D level. Compounds canbe encoded by physiochemical properties (1D), structural arrangement (2Dgraphs), as well as computed features relating to 3D.  We want to find ways that chemical representation can enhanceabilities of existing kernel functions? Additionally, we explore kernels or methods fit for chemical compoundmanipulation so 1D, 2D, 3D, and 4D molecular information are used to optimallylearn information to improve the drug discovery process.   As an initial step, to addressthe encoding question, rCDK in R is being used to attain fingerprints. R isalso used for necessary statistical analysis and machine learningmethods(kKmeans, feature extraction/selection, and SVM/LS-SVM). Overall, we aredeveloping a platform for innovative chemoinformatics for use with various drugdiscovery related applications

Chemoinformatics in R: Enriching molecular encoding, selective use of kernels and methods

239th National Meeting (2010)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.