Biological information
management and analysis
as illustrated by malaria research
1. Problems
2. Managing data context
3. Managing and analyzing data
Factors in combating malaria
Scientific: biology, ecology, chemistry, etc.
EnvironmentalCultural/Sociological
Economic Political/Ethical
Scientific layers
Psychology/ Emergent properties of brain andSociology populations
Biology All complexity of element interactions(macromolecules, cells, brain, populations)
Chemistry Properties of "simple" element interactions
Physics Properties w/o inter-element interactions
The labyrinth of biological research
- Which direction to follow and in what way?
- What relevant information is available?
- How to keep a good record of the path?
- How to find useful collaborators?
- What do the results imply?
Researchers are drowning in the sea of information…
Problems with "physicalization" of biology
• Data richness
• Data sharing and integration
• Model-data correspondance
• Understanding bioresearch problems
• Understanding bioresearch constraints
• Tim Berners-Lee, James Hendler. Scientific publishing on the 'semantic web'. Nature Debates, April 2001.
• Jonathan Knight. Negative Results: Null and void. Nature, April 2003.
• Who'd want to work in a team (Editorial). Nature, July 2003.
• Declan Butler. Open-access row leads paper to shed authors. Nature, September 2003.
Information problems in the Nature journal
Information management needs of Anopheles GPH
• Inform scientific community (publications, database submitions, conferences…)
• Prevent loss of information (unpublished results, method details, …)
• Report to administration (progress, problems, …)
• Share and manage supplies (materials, equipment, …)
• Share informational resources (protocols, bibliography, …)
• Facilitate collaboration (share information, co-author documents, …)
IP, France IP, Korea
IP, SenegalIP, Madagascar
Columbia University, USA
(outside collaboration)
… … … …
Permanent, shared information (< 30%) :
Temporary, individual information (100%) :
Sources of Research Information: Status quo
Journals
NotebooksComputer Files
Databases
Permanent, shared information (100%):
Sources of Research Information: Ideal
Integrated Repositories of Structured Data
1. Problems
2. Managing data context
3. Managing and analyzing data
Researcher
AdvisorAdministration
Scientificcommunity
CollaboratorsResearch
group
Flow of research information: at present
Flow of research information: proposed
AdvisorAdministration
Scientificcommunity
CollaboratorsResearch
group
Researcher
Database
Unstructured information:
Research notes, Contracts, Project reports, Clinical trials documentation …
2 types of information
Structured information:
GenBank, Medline, Employee database, Invoice database, …
Forms
Documents
Methods of contributing written information
• Traditional documents - hard to search and manipulate
• Traditional forms - overly constraining, hard to create documents
• Structured documents (New!) - best of both worlds
Malaria surveys were carried out in two rural villages near the town of Ziniaré (35 km northeast of Ouagadougou) in a shrubby savanna of the Mossi plateau . An
intense P. falciparum transmission is detected …
Problems with forms
Project: Measurements of response to …
Experiment: Entomological Observations of …
The ability to resist Plasmodium falciparum malaria is an important adaptive trait of human populations living in …
The results of our comparative study show consistent interethnic differences in P. falciparum infection …
Method: Observations
Different response to P. …
Summary 1
• Biological function is based on infinity of interactions between basic elements
• Biologists are drowning in the complexity of information
• Need to understand biological problems and constraints before applying analytical approaches
• Need to resolve the problem of information storage and retrieval
Form constraints
1. Limited number of categories
2. Limited number of fields per category
3. Constrained field space
4. Limited editing (copy, move, delete, etc.)
5. No coherent document representation
6. Unable to represent complex hierarchical information
Database
iPad middle-layerserver
"3-tier" architecture of the iPad system
iPad EditoriPad Web Portal
iPad Demo
Major Benefits
Monetary savings:+ Less lost work+ Resource optimization
Time savings:+ Faster search+ Faster communication and formatting+ Less lost work
Increase in the quality and quantity of research:+ Useful perspectives+ Improved collaboration+ Improved project management+ More information given to the Institute community+ More information given to the scientific community (in the future)+ A tool to structure scientific data (in the near future)
Drawbacks
- Learning new software (very simple)
- Changing habits (will go away over time, gradual adoption)
Support for structured documents
1. WWW Consortium, industry analysts
2. General systems within the past year
(Microsoft, Arbortext, Altova, etc.)
3. Specific systems in the military
Evolution of information (Tim Berners-Lee)
First Consulting Group, "XML and Pharmaceutical Industry" (2003) :
"In order to be profitable and competitive as they serve our global healthcare needs, drug companies require information systems to help them work efficiently to deliver a high-quality product. With that in mind, momentum is growing to leverage XML technology in the content management and publishing systems, being used by the pharmaceutical industry throughout the drug development lifecycle."
* Interest from Aventis Pharma, Sopra Group, Genset
Gilbane Report, "XML for Content" (2003):
"So what's the biggest problem with XML content? Authoring it… The authoring tools are becoming more capable and people are starting to figure out that the ease of processing XML content can outweigh the pain of creating it, but there is still some way to go."
1. Problems
2. Managing data context
3. Managing and analyzing data
Summary 2
• Data context is important both for information management and for data interpretation
• iPad can be used to structure data context using XML markup
• Structuring data context is the precursor for better structuring of data.
3 Steps to "Paradise"
1. Agree on standard organizational categories
SB-UML
Gene Ontology Bioprocess ontology …
"Dynamic" ontologies
Bioprocessontology
3 Steps to "Paradise"
1. Agree on standard organizational categories- "Dynamic" ontologies, Gene Ontology, Bioprocess ontology, …, SB-UML.
2. Sort information into the ontological categories- Data mining algorythms, Electronic forms, Semantic markup.
<protein>p53</protein><interaction>activates</interaction><gene>CD95</gene>
Dynamic ontology
Entity Property Relation
BioStructure Process Data Method
Molecule MolecularComplex Organelle Organ Tissue Organism
name
alternative names
type
value
Data markup
X protein activates Y gene in A. gambiae salivary glands.
Molecule (name: X, type: protein)
Molecule (name: Y, type: gene)
Relation (name: activates, type: molecular interaction)
Entity (name: A. gambiae, alt. name: Anopheles gambiae, type: organism)
Entity (name: salivary glands, type: organ)
3 Steps to "Paradise"
1. Agree on standard organizational categories- "Dynamic" ontologies, Gene Ontology, Bioprocess ontology, …, SB-UML.
2. Sort information into the ontological categories- Data mining algorythms, Electronic forms, Semantic markup.
3. Develop search, visualization, and analysis tools- Blast, Bioprocess and molecular modeling, Concept network, …
Concept node
- Better global picture to see where to go
- Helpful info along the way
- Organized research process
- Better ways to share data
- Broader impact of results
- Modeling and simulation tools
Summary 1
• Biological function is based on infinity of interactions between basic elements
• Biologists are drowning in the complexity of information
• Need to understand biological problems and constraints before applying analytical approaches
• Need to resolve the problem of information storage and retrieval
Summary 2
• Data context is important both for information management and for data interpretation
• iPad can be used to structure data context using XML markup
• Structuring data context is the precursor for better structuring of data.
Summary 3
• 2 steps for structuring data: ontology + methods for data entry
• Simple "dynamic" ontologies can be used to derive standard "static" ontologies
• iPad-like system can be used to simplify structuring biological data
• Data analysis, modeling, and simulation tools need to be data-driven, generic, and easy to use.