University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program

title

Visualisation and data integration for genomics

speaker

Dr. Ela Hunt, ETH Zurich, Switzerland
Zürich, Switzerland

date & place

Wednesday, 20.06.2007, 16:15 h
Room C252

abstract

(in collaboration with Joanna Jakubowska and Matthew Chalmers at Glasgow)

VisGenome is an interface to genome data from Ensembl, www. ensemble.org. It was developed to address data analysis needs of our biomedical collaborators who search for genes that cause a variety of diseases. One of the methods of work is to view data on maps at Ensembl and compare them with the results of experiments carried out locally. Usually, a biologist looks at an area containing 100 to 200 genes, and views their private experimental results which show some of those genes as being active or not active, in healthy and sick individuals. At the same time the biologist wants to see the corresponding gene area in related species (mouse, rat, human) and see if the corresponding genes are known to be involved in similar disorders.

It is impossible to get a clear visual overview of 100 or 200 genes in Ensembl, as the web page often crashes when so much data is being retrieved, and the page cannot clearly show so many genes at once. It is also impossible to visually compare such a number of genes in two species and verify their relationships. We hypothesised that by providing improved zooming, ordering genes vertically, which allows for different label placements, and providing views of gene relationships, we may be able to improve support for the data analysis tasks our collaborators face. To verify this hypothesis, we carried out a user study with 15 users. The study included three object and interval localisation tasks. The study shows that the users are more successful in performing the same tasks in VisGenome than in Ensembl, but the time required to carry out the task is similar.

The next stage in the investigation involves the superposition of experimental data derived in the lab or gathered from a journal publication onto the map. This has been done programmatically for several data sets and proved to assist research. We are now building a generic mechanism for data superposition, using the metaphor of a map mashup. Our mashup uses GeneCards (2GB of XML) as an example data set from which the user can select data to be added as object annotations or new objects.

I will discuss this work, and then provide a short demo of VisGenome.