VisGenome and Ensembl: Usability of Integrated Genome Maps

1
VisGenome and Ensembl Usability of Integrated
Genome Maps

• Asia Jakubowska1
• Ela Hunt2, Matthew Chalmers1, Martin McBride3 and
  Anna F. Dominiczak3
• 1Department of Computing Science, University of
  Glasgow, UK 2Department of Computer Science, ETH
  Zurich, Switzerland
• 3BHF GCRC, University of Glasgow, UK
• http//www.dcs.gla.ac.uk/asia

2
Contribution

• It is not always clear how best to represent
  integrated data sets, and which application and
  database features allow a scientist to take best
  advantage of data coming from various information
  sources.
• Solution user study comparing an existing
  general-purpose genetics visualisation and query
  system, Ensembl, to our new application,
  VisGenome.

3
Database integration and usability

• Database and visualisation usability studies
  guarantee user satisfaction.
• User-centered design for database query systems
  (Catarci, Inf. Syst., 2000).
• Database usability issues creation and use by
  end-users (Jagadish, SIGMOD, 2007).
• Reducing visual overload in VisGenome (selected
  data types, panel info with additional
  information) supports the process of data
  analysis and contextualisation.

4
Motivation for visualisation research

• Growing amounts of biological data (millions of
  objects on a chromosome)
• Cognitive problem how to understand the data
• How to compare data coming from various sources
  and experiments
• Solution visualisation a new genome browser,
  VisGenome with a novel algorithm, and user
  studies
• Methods user studies, evolutionary design

5
Searching for genes responsible for hypertension

• Target users biological research groups working
  in cardiology, metabolic diseases, schizophrenia
  and cancer
• User activity animal breeding, genotyping and
  µArray experiments (analysing the activity of
  genes, including faulty genes), human genotyping
  and µArrays, proteomics
• Data interpretation in a browser
• Rat breeding visual interpretation of results,
  for each subtype of rat

6
Goal finding a gene which causes disease, by
eliminating the contribution of other genes
(minimising size of the area under investigation)
7
Gene browsing tools

• Chromosome 100-290 mln letters
• QTL chromosomal region associated with a
  disease 3-50 mln letters
• Gene 20-70 kb letters
• Microarray probe 25 letters
• SNP (Single Nucleotide Polymorphism) 1 letter

8
Data our Biological Collaborators Use

• The mouse, the rat and the human chromosomes
• Details for QTLs, genes, microarray probe sets,
  SNPs
• Data for animals the biologists experiment with
  and at the same time data for humans (patients)
• Comparison between three species and various rat
  strains at the level of QTLs, gene order,
  microarray results and gene sequences

9
VisGenome - goals

• A legible display for both LARGE and SMALL
  objects, in one screen (QTLs, genes, micro array
  probes, markers)
• Context and detail
• Smooth zooming, panning, scaling
• Client-side view adjustment and data exploration

10
VisGenome
11
Ensembl www.ensembl.org

• The most common tool among our collaborators
• An interactive Web site, using server-generated
  images
• Genome browser in SyntenyView
• Clickable pictures

12
Ensembl browser
context

• Labelling and searching
• There is no possibility of displaying all parts
  of the view on one page (three screens are
  needed)
• No user manipulation of the image

13
Initial quantitative user study

• Tests (3 tasks) with the hypertension and
  schizophrenia research groups
• Questions about visualisation techniques and
  their usability during biological work
• Measuring the time and mouse clicks
• GUI screen recording
• Feeding into new VisGenome versions

14
Experiment overview

• Participant Consent Form
• VisGenome present the tool
• VisGenome an example task
• VisGenome the 3 experimental tasks
• VisGenome Workload Tests (NASA TLX)
• Ensembl the 3 experimental tasks
• Ensembl Workload Tests (NASA TLX)
• Questionnaire

15
Experiment

• 15 participants - 6 Ensembl Experts and 9 Non
  Ensembl Experts
• 40 min the quickest user
• 2 hours the most sociable user
• almost 4 months elapsed time
• 1st user took part on 11.08.2006, 15th
  01.12.2006
• user errors specific to tool and task

16
Tasks

• Showing the first gene and the last Affymetrix
  probe inside the specified region in the Single
  Representation.
• Finding homologous genes between the rat
  chromosome 18 and the human chromosome 5 in the
  Comparative Representation.
• Finding the longest QTL and the genes which are
  the closest to the beginning and the end of the
  QTL in the Single Representation.

17
Interview goals

• What data do the users expect to see?
• Information presentation colour, font, layout
• Display legibility
• Aesthetic appeal
• Ease of use zooming, scaling, interactivity
• Effectiveness
• NASA TLX measure of workload, etc.

18
Additional interview Questions - results

• Most people knew about zooming and panning
• A few knew about Fisheye and Excentric labeling
• Most preferred zooming via mouse action
• But 2 strongly prefer buttons
• 8 users wanted different colours, 7 do not mind
• Karyotype representation preferred by users
• 8 users preferred dynamic info panel to all data
  in one view
• Most preferred clicking for choosing chromosomes,
  only 1 preferred dragging

19
Karyotype representations
20
Task completion with correct answers
21
Task completion (imprecise execution)
22
Workload tests (NASA TLX) results
23
Summary of the initial quantitiative user study

• Tasks with Ensembl were less successful
• Typing coordinates with mistakes
• Showing the longest QTL in a small region
• VisGenome advantages supports comparison of
  distant details
• Confused by the use of the mouse for zooming, but
  got used to it quickly
• Centring is needed once an object is selected if
  it is to stay in the centre during zooming

24
Talk Summary

• Motivation information overload and need for
  experimental assessment of usability
• VisGenome and Ensembl
• Initial quantitative user study
• Findings neither VisGenome (first version) nor
  Ensembl are perfect for the users and the users
  made a number of mistakes during the experiment.

25
Questions ?

• http//www.dcs.gla.ac.uk/asia/VisGenome
• (jar files, source code, user handbook)
• asia_at_dcs.gla.ac.uk

Matthew Chalmers
Ela Hunt
me
Anna Dominiczak
Martin McBride