Applied bioinformatical visual analytics

In modern biology and medicine, data is often analysed in a visual way. In genetics, for example, disease inheritance patterns are investigated based on drawn pedigrees. However, traditional information visualisation methods cannot cope with the demands arising from the large amounts of heterogeneous, multi-variate data sets that have come to characterize research in modern life science. Furthermore, the integration of data analysis methods in the visualisation process and the support of explorative data analysis have become critically important undertakings.

This thesis highlights how the knowledge discovery process in life science research can be supported by the tight combination of visual and analytical methods. We show how this principle of "Visual Analytics" can be applied to tackle the corresponding bioinformatics challenges, such as: large mass of heterogeneous data, complex exploratory tasks, and multidisciplinary teams. For this purpose, we developed in two different biological domains tools following the "Visual Analytics Mantra": PedViz for the identification of disease clusters and pedigree splitting and TisViz for the discovery of proteomic biomarkers. Moreover, we performed an extensive evaluation of our tools to benchmark their computational performance and to investigate the value of a visual analytical approach to answer biological research questions. Results indicate that our domain specific approaches provide additional insight into the data and that with the simple use of pure visualisation or pure data analysis methods, or combinations of the two, not the same level of comprehension can be achieved.