Strategies and Tools for Multivariate Biology: Tackling High Dimensional Postgenomic Data

Postgenomic biology is characterised by large and diverse datasets presenting many technical challenges in data handling, analysis and integration. Further, there are significant challenges in converting the data into biological knowledge to support model building and hypothesis generation. A basic strategy is to find structure in the data (clusters) representing the coordinated response of system probes (e.g. biomolecules) to given stimuli. Clusters may then be characterised by reference to knowledgebases (such as pathway genome databases). A complementary strategy is to select probes based on extant knowledge (e.g. related by terms in an ontology) and explore the character of their biological response. We have developed an interactive software tool, Vitamin-B (visual, interactive tool for the analysis and mining of bioinformatics data), which provides utilities for both supervised and unsupervised inquiry and links to external resources to promote knowledge acquisition in support of model construction and hypothesis generation. The focus in the tool´s design and development is on usability: aiming to lower the barriers for both expert and novice users to interactive exploratory and targeted data analysis through a full complement of analysis methods coupled to linked dynamic graphics. The tool is based on the proven components of R and ggobi and provides enhanced functionality through a custom user interface, flexible data selection, linkage to external resources, undo functionality and the production of summary reports. A key feature of the project is the central role of HCI principles in the interface. Evidence for the effectiveness of the tool (and its components) will be acquired by user evaluation in illustrative goal-driven scenarios.