The Connectivity Map has introduced an approach to discovering and elucidating connections between disease, drugs, and genes. This collection of transcriptional expression data has been used to recover known relationships between drugs and disease states as well as providing a basis for new ones. Examining small molecule compounds with the Connectivity Map can reveal information about the action of those compounds. Some of the compounds that we studied form well-established pharmacological classes, but a number of them are less studied in the literature. In order to further explore relationships among small-molecule compounds, we examined approximately 3,000 bioactive compounds. We began by using summarized connectivity, an indicator of the strength of functional relationships, to measure similarity between compounds. To select a hierarchical clustering model, we employed a consensus clustering algorithm that determines a clustering of rank k that is robust to resampling. We then compared these clusters to a set of known pharmacological classes. Preliminary results indicated a number of robust clusters, including some novel clusters, that may be of interest. Further analysis and validation of the clusters may offer some insights about the pharmacological and biological functions of these compounds.
PROJECT: Unsupervised discovery of functional classes of small-molecule compounds using the Connectivity Map
The Broad is a unique place with a diverse array of research areas and scientists. I learned to view and communicate science in new ways and gained an invaluable group of peers and mentors. The Broad experience changes you for the better and prepares you to join the next generation of researchers.