Bozeman, Montana (August 6, 2002) - ChemTree® sequential screening software has been upgraded to include a new distance metric dubbed Grand Unified Compound Clustering Intelligence (GUCCI). With this feature, chemists can build 1000's of recursive partitioning tree-based models all driven by the specific structural features that enhance or reduce a compound's activity, and distill these models into a single visual display. The GUCCI all-compound-pairs distance metric is based on the idea that when two compounds end up together in a small subset deep with within a tree, the structural features that drive their activity are shared. Hence, the compounds are similar in activity space. Using multiple trees as a panel of experts to generate an all-pairs distance matrix in structure/activity space, which when ordered using principle component analysis, gives a unified picture of how the forest of trees are collectively clustering the compounds.
"This will significantly enhance the ability of scientists to pull apart high throughput screening data sets," says Christophe Lambert, Ph.D., CEO and Founder of Golden Helix, Inc. "This feature solves the problem of 'what's the one true model?'. Traditional methods that cluster compounds by their similarity tend to not be able to weight which parts of the compounds matter." By combining recursive partitioning with this distance metric clustering, the ChemTree GUCCI feature can cluster compounds according to similarity of structure, activity and binding mechanism all at once. This allows compounds that may be extremely different according to subgraph comparisons to be grouped together if they share a key functional group that drives activity. This has tremendous implications for sequential screeners and medicinal chemists who can now take apart a large data set in a way that is more meaningful to the assay in question. An upcoming multivariate version of this methodology will allow clustering according to multiple targets - for instance coupling together parallel screens of efficacy, safety, and specificity.
"ChemTree software can increase your hit rates by an order of magnitude while reducing the number of compounds screened by 80%," reports Dr. Lambert. ChemTree is statistical analysis software that facilitates intelligent, target specific compound selection for High Throughput Screening (HTS). ChemTree takes the compound structures and results from a screen, generates a model that predicts results based on molecular features and applies that model to untested compounds, thereby increasing hit rates on future testing. The underlying technology behind ChemTree is recursive partitioning customized for sequential screening.
About Golden Helix, Inc.
Founded in 1998 in Bozeman, MT, Golden Helix is a leading provider of data analysis software products and services designed to accelerate life science research and increase pharmaceutical R&D productivity. The company's products enable hundreds of researchers at the world's top pharmaceutical, biotech and non-profit research organizations to discover the genes that cause disease, improve target and lead discovery processes, and advance pharmacogenetic initiatives. Its products have been cited in nearly 100 peer reviewed publications that detail ground-breaking research for discovering the genetic and environmental basis of disease and creating safer and more effective medicines.
Golden Helix is committed to Accelerating the Quest for Significance.
For more information, please visit www.goldenhelix.com.
Director of Marketing
Golden Helix, Inc.