Dr Alan Hogg

1
The Complexity of Informatics

• Dr Alan Hogg
• Section Head - Platform Development
• NCRI Informatics Coordination Unit

2
Aim

• What are researchers trying to do?
• How can informatics support that aim?
• What are (some of) the complexities that
  challenge us?

Looking at Informatics as a whole, not just
specifically ONIX the ICU
3
Research Scenario
Drug X is an effective treatment for 10 women
with breast cancer What are the molecular
signatures associated with efficacy?
Prof Herbie Newell Former Director of
Translational Research at CR-UK
4
Research Steps

• Create Clinical Data Set

• Online access to original patient tissue samples
  from tissue repository

• Microarray analysis to sub-type tumours

• Identify molecular pathways associated with
  candidate genes

• Qualify biomarkers with more samples and in
  prospective clinical trials

5
Research Outcome

• Biomarkers that can identify women with breast
  cancer who are most likely to respond to Drug X
• Better understanding of metabolic pathways
  involved
• New potential drug targets for future studies

6
The Complexity of Informatics?

• Cuts across scientific domains -
• Clinical trials, genomics and systems biology
• Each is a discipline with their own terminologies
• Different forms of information (data types) -
• Clinical records, images, genomics literature
• Different processes -
• Extracting existing data (clinical information)
• Generating and analysing new data (genomics)
• Searching literature (metabolic pathways)

Without Informatics this is NOT possible,
however, it is not easy either
7
1. Clinical Data Set

• Willingness to share data
• Data sharing policies in place with sponsors of
  research
• Privacy policy where sensitive data exists
• Public confidence in data privacy
• Informatics needs to support those privacy
  policies
• Support authorisation and authentication
• Secure transfers between networks and
  institutions (not a CD in the post!)

8
2. Online access to original patient tissue
samples from tissue repository

• Data Linkage
• Ability to link between data sets without
  compromising privacy
• HIRU in Wales have developed the ALF system
• Prevent identification through access to certain
  combinations of data
• GIMI project at Oxford Computing Laboratories
• Anonymisation vs Pseudo-anonymisation
• Support for Honest Brokers

9
2. Online access to original patient tissue
samples from tissue repository

• Data Standards
• Clear meta-data (data about data)
• For example defining the scale of the images,
  date image taken etc
• Pathology image annotations need to have a clear
  and consistent meaning
• RADLEX new pathology annotation standard being
  developed in the US
• Informatics needs to support the creation and
  easy access to such standard vocabularies
• Development and sharing of tools to support
  annotation

10
3. Microarray analysis to sub-type tumours

• Automation - to collect data output and pass it
  through a suitable analysis tool
• Requires physical interoperability
• Issues with interfaces (access points) not
  existing
• Consistent meaning of information is vital
• Again promotion of standards
• Initiatives such as NCIs caBIG, EMBL-EBI
  EMBRACE, and ONIX trying to address this

11
3. Microarray analysis to sub-type tumours

• International Repository to facilitate data
  sharing by putting all the data together
• Variation in data definitions and formats a
  hindrance
• Need professional data managers to curate it
• Physical infrastructure would be large
• Current funding models in UK not designed to
  support long-term repository programmes

12
4. Identify molecular pathways associated with
candidate genes

• Knowing what exists
• What stuff exists that might be useful to you
• Resource catalogues simply knowing who has
  what and where.
• ICU working with NHS Information Centre, Research
  Capability Programme, NCIN and caBIG to share
  info.
• Ability to search literature and data
  repositories
• Biologically Aware searching across multiple
  repositories is Holy Grail
• ICU collaborating with UCL, EBI and caBIG

13
5. Qualify Biomarkers

• Collaboration between teams
• Physical and political ability to operate
  across project teams, institutions and national
  boundaries
• Publication of results
• British Library (publications) leading a project
  with EMBL EBI (raw data) and Nature (metabolic
  pathways) how do we maintain these links?

14
Informatics Challenge - Summary

• Technical
• Physical ability to share data between networks,
  individual systems and research teams
• Data Definitions are clear and meaningful
• Meta-data its units of measure etc
• Data Standards to allow clear, consistent
  annotation classification of research findings
• Data Relationships defined linkages between
  data sets
• Hearts and minds
• Willingness to share
• Informatics systems must support the policies
  under which the work is done

15
Thank You
16
Backup slides
17
The Landscape

• Scientific Domains
• Revealed a diverse set of both practical/experimen
  talist domains
• An equally diverse landscape of
• Data capture and analysis
• Storage
• Maintenance and curation
• Formats and standards
• IT infrastructures