Thumbnails:
List:
Year:
Category:
Session:
Poster:
Getting poster data...
A Heilbut, X Zhu, P Tamayo, ME Hatten, E Kolaczyk, JP Mesirov (7 Cambridge Center, Cambridge, MA)
Longitudinal gene expression profiles are often measured over multiple experimental contexts such as cell types or drug treatments, and unsupervised clustering is used to discover patterns. When experiments are organized in blocks corresponding to specific contexts, one would like both to distinguish clusters within blocks (temporal patterns in one cell type) and to explore how clusters relate to each other across blocks. BOMBASTIC is a simple tool for interactive exploration of blocked, clustered data. Given an analysis specified as a chain of clustering steps and filters, BOMBASTIC generates a program to cluster each block independently and an HTML5 interface to visualize and prune subsets from the implied tree of cluster assignments. Distributions of model statistics (eg. fit, scale, intercept) relating individual points to clusters are shown as 1-D or 2-D histograms on which selections can be defined to cross-filter data of interest. We demonstrate application of BOMBASTIC to analysis of cell-type specific gene expression profiles of developing neurons and differentiating embryonic stem cells.