data integration issues in biodiversity research jessie kennedy shawn bowers, matthew jones, josh...

Data Integration Issues in Biodiversity Research

Jessie KennedyShawn Bowers, Matthew Jones, Josh Madin,

Robert Peet, Deana Pennington, Mark Schildhauer, Aimee Stewart

Visual Tools for Managing Taxonomic Concepts

SEEKScience Environment for Ecological Knowledge

Research and develop information technology to radically improve the type and scale of ecological science that can be addressed

Visual Tools for Managing Taxonomic ConceptsBiochemistry

Climatology

Taxonomy

Meteorology

Nomenclature

Paleontology

GenomicsProteomics

Hydrology

Morphology

Geology

Oceanography

Geography

Ecology

Science and Scientific Data are Complex


Biochemistry

Climatology

Taxonomy

Meteorology

Nomenclature

Paleontology

GenomicsProteomics

Hydrology

Morphology

Geology

Oceanography

Ecology

Geography

Organism

Name

Taxon concept Gene

sequence

Pathway

Protein

Location

TemperatureDepth


Individual Scientist

Small Scientific Community

Large Scientific Community Scientific Laboraotory

Scientific Community: complex


Biochemistry

Climatology

Taxonomy

Meteorology

Nomenclature

Paleontology

GenomicsProteomics

Hydrology

Morphology

Geology

Oceanography

Ecology

Geography

Organism

Name

Taxon concept Gene

sequence

Pathway

Protein

Location

TemperatureDepth

Biochemistry

Climatology

Taxonomy

Meteorology

Nomenclature

Paleontology

GenomicsProteomics

Hydrology

Morphology

Geology

Oceanography

Ecology

Geography

Organism

Name

Taxon concept Gene

sequence

Pathway

Protein

Location

TemperatureDepth

Biochemistry

Climatology

Taxonomy

Meteorology

Nomenclature

Paleontology

GenomicsProteomics

Hydrology

Morphology

Geology

Oceanography

Ecology

Geography

Organism

Name

Taxon concept Gene

sequence

Pathway

Protein

Location

TemperatureDepth

Biochemistry

Climatology

Taxonomy

Meteorology

Nomenclature

Paleontology

GenomicsProteomics

Hydrology

Morphology

Geology

Oceanography

Ecology

Geography

Organism

Name

Taxon concept Gene

sequence

Pathway

Protein

Location

TemperatureDepth


Science & Scientific Data are Continually ChangingConclusions become

foundations for new hypotheses

New experiments invalidate existing knowledge

Knowledge is open to interpretation Different opinions

Need to build this into our technological solutions

observation

experiment hypothesis

conclusion


Exploiting Scientific Data

To support scientists in Discovery Access Sharing Integration/Linking Analysis

Scientists can then improve their potential for new scientific discovery


Data Integration/Linking: approaches

Metadata to describe the data sets and know how to interpret the data sets

Ontologies to define the terminology used and know how data might be related and

to aid automatic transformation of the dataStandardisation of formats

for exchange of data + to ease integrationLSIDs

to uniquely identify things; know when 2 things are the sameWorkflows

to enable specification, refinement and repetition of integration/analysisProvenance of data

to record where the data has come from and what has happened to it en route.


Projects in most sciences:

ESG


Ecological Science - AnalysisEcological niche modeling of species distributions

Where do species occur now?

Image from http://www.lifemapper.org

Where will they occur in the future?


Ecological Niche Modeling

EnvironmentalCharacteristicsfrom griddedGIS layers

Known Species Locations

Temperature layer

Many other layers

EnvironmentalChange

Prediction

Future ScenariosOf EnvironmentalCharacteristics

InvasionArea

Prediction

EnvironmentalCharacteristicsOf Different

Geographic Area

NativeDistributionPrediction

EnvironmentalCharacteristics

Of Surrounding Geographic Area

DevelopModel

MultidimensionalEcological Space

D1 = Temperature

D2 Dn


Sources of Scientific Data

Data are massively dispersed Ecological field stations and research centers (100’s) Natural history museums and biocollection facilities (100’s) Agency data collections (10’s to 100’s) Individual scientists (1000’s)

Data are heterogeneous Syntax (format) Schema (model) Semantics (meaning)


Challenge: Data Integration


SEEK Components


Semantic Annotation – SEEK ontologies

Integration/merge Concept mapping Units conversion Spatial & temporal scaling

Data discovery Finding relevant data sets Understanding data set content


Smart (Data) Integration: MergeDiscover data of

interest

… connect to merge actor

… “compute merge”


Smart Merge …Semantic type

annotations and ontology definitions used to find mappings between sources

Executing the merge actor results in an integrated data product (via “outer union”)

a1 a2 a3 a4a 5 10b 6 11

a1 a2 a3 a4a 5 10b 6 11

a5 a6 a7 a8 0.1 a 0.2 c 0.3 d

a5 a6 a7 a8 0.1 a 0.2 c 0.3 d

a3

a6

a1

a8

a4

Merge

a1a8

a3a6

a4

BiomassBiomass

BiomassBiomassSiteSite

SiteSite

a1 a3 a4a 5.0 10b 6.0 11a 0.1c 0.2d 0.3

a1 a3 a4a 5.0 10b 6.0 11a 0.1c 0.2d 0.3

Merge Result


Challenges of Taxonomic Data

Scientific names change in meaning over time + geographical region conclusions being drawn from analysis of data integrated on names.


Flora North America

SubAlpine Fir

USDA Plants & ITIS

Abies lasiocarpa

Abies bifolia

Abies lasiocarpa

var. arizonica

var. lasiocarpa

What is Abies lasiocarpa?


Aus L.1758

Aus aus L.1758

Linneaus 1758

Aus aus L.1758

Tucker 1991

Aus L.1758

Aus cea BFry 1989

Aus aus L.1758

Aus L.1758

Aus bea Archer 1965

Archer 1965

Aus aus L.1758

Aus L.1758

Aus bea Archer 1965

Aus cea BFry 1989

Fry 1989

Aus L.1758

Xus beus (Archer) Pargiter 2003.

Aus ceus BFry 1989

(vi) Xus Pargiter 2003

Pargiter 2003

Aus aus L. 1758

Changes in meaning of names

Aus bea and Aus cea noted as invalid names and replaced with Aus beus and Aus ceus.

Pyle 1990

5 Revisions of Aus 1 name spelling change

Taxonomic history of imaginary genus Aus L. 1758


Aus L.1758

Aus bea Archer 1965

Aus aus L.1758

Archer 1965

Aus L.1758

Aus aus L.1758

Linneaus 1758

Aus aus L.1758

Aus L.1758

Xus beus (Archer) Pargiter 2003.

Aus ceus BFry 1989

(vi) Xus Pargiter 2003

Pargiter 2003

Aus aus L. 1758

Aus bea and Aus cea noted as invalid names and replaced with Aus beus and Aus ceus.

Aus aus L.1758

Tucker 1991

Aus L.1758

Aus cea BFry 1989

Aus L.1758

Aus bea Archer 1965

Aus cea BFry 1989

Fry 1989

Changes in meaning of names

Pyle 1990

• 8 Names• 2 genus• 6 species

N4 - Aus beus Archer 1965

N1 - Aus aus L.1758

N1

C1.5

C1.4

C1.3

C1.2

C1.1 C1.1 - Aus aus L.1758 sec. Linneaeus 1758

C1.2 - Aus aus L.1758 sec. Archer 1965

C1.3 - Aus aus L.1758 sec. Fry 1989

C1.4 - Aus aus L.1758 sec. Tucker 1991

C1.5 - Aus aus L.1758 sec. Pargiter 2003

N2 - Aus bea Archer 1965

N5 C5.5N5 - Aus ceus Fry 1989

C5.5 - Aus ceus Fry 1989 sec. Fry 1989

C6.5N6N6 - Xus beus Pargiter 2003

C6.6 - Xus beus Pargiter 2003 sec. Pargiter 2003

N2

C2.3

C2.2 C2.2 - Aus bea Archer 1965 sec. Archer 1965

C2.3 - Aus bea Archer 1965 sec. Fry 1989

N3

N4C3.4

C3.3N3 - Aus cea Fry 1989 C3.3 - Aus cea Fry 1989 sec. Fry 1989

C3.4 - Aus cea Fry 1989 sec. Tucker 1991

N0 - Aus L.1758

N0

C0.5

C0.4

C0.3

C0.2

C0.1 C0.1 - Aus L.1758 sec. Linneaeus 1758

C0.2 - Aus L.1758 sec. Archer 1965

C0.3 - Aus L.1758 sec. Fry 1989

C0.4 - Aus L.1758 sec. Tucker 1991

C0.5 - Aus L.1758 sec. Pargiter 2003

C7.5N7

N7 - Xus Pargiter 2003

C7.6 - Xus Pargiter 2003 sec. Pargiter 2003

8 Names 17 Concepts

Each name has many

concepts ormeanings


Find data sets containing Aus aus

Many possible interpretations of Aus aus (N1)

Original concept: C1.1 Most recent concept: C1.5 Preferred Authority (e.g. Fry 1989): C1.3 Everything ever named N1:

Union(C1.1,C1.2,C1.3,C1.4,C1.5) Best fit according to some matching algorithm

Best(C1.1,C1.2,C1.3,C1.4,C1.5) New concept containing only those features

common to all concepts with the name N1: Intersection(C1.1,C1.2,C1.3,C1.4,C1.5)

Is it appropriate to link or merge data sets returned on the scientific names? Depends on the user’s purpose Level of precision required

N1 - Aus aus L.1758

N1

C1.5

C1.4

C1.3

C1.2

C1.1


C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N0 N7

N1 N2N5 N6N3 N4

Information from literature on synonymy

Taxonomists record which names their concepts are synonymous withand any name changes

Parent child relationships in 5 revisions

Names for each of the concepts


Find data sets with Aus aus (N1)

C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N0 N7

N1 N2N5 N6N3 N4N1

C1.1 C1.2 C1.3 C1.5C1.4

N1



C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N0 N7

N1 N2N5 N6N3 N4N1 N2

C1.1 C1.2 C2.2 C1.3 C2.3 C1.5C1.4

N1



C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N0 N7


C1.1 C1.2 C2.2 C1.3 C2.3 C1.5C1.4 C3.4C3.3

N1 N2N2 N3



C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N0 N7


C1.1 C1.2 C2.2 C1.3 C2.3 C1.5C1.4 C3.4C3.3 C6.5

N6N3 N4N1 N2N2



C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N0 N7


C1.1 C1.2 C2.2 C1.3 C2.3 C1.5 C5.5C1.4 C3.4C3.3 C6.5


Results in everything returned for Aus aus by traversing the synonymy and name links


C1.5 C5.5

C0.5

C1.4 C3.4

C0.4

C1.1

C0.1

C1.2 C2.2

C0.2

C1.3 C2.3 C3.3

C0.3

C6.5

C7.5

N1N5 N6

N2 N3 N4

N0 N7

= =

Information to improve data sets returned

Minimally what we need are set relationships from concepts in any taxonomy to earlier concepts

and name changes related to earlier names

We can build systems to return data suit for purpose


Real Biological TaxonomiesLarger and change more frequently than the Aus exampleGerman mosses

14 classifications in 73 years covering 1548 taxa only 35% thought to be stable concepts

65% of names used in legacy data sets are ambiguous

Taxonomic Revisions of genus Alteromonas 34 years: from 1972 to 2006 At the species level

18 “emendations” 19 species reassigned to 4 genera

3 new combinations 6 synonyms 2 species to subspecies 2 subspecies to species

21 new species


SEEK Taxon ApproachUse Taxon Concepts for referring to organisms

Aus aus L. 1758 sec. Tucker 1991 Abies lasiocarpa (Hook) Nutt. sec FNA 1997

Taxon Concept/Name Resolution International data exchange schema

TCS (Taxonomic Concept Schema)

Concept Repository and Resolution web service Linked to Kepler workflow system Globally unique identifiers (LSIDs) Visualization software for comparing Taxonomies and

Asserting Concept Relationships


Taxon Object Server

Mammal Species of the World

TaxonomicLiterature

TaxonomicData

Providers

TOS

SEEK Cache

Databaseto TCSMappingTool

ConceptExtractionTool

TCS

TCS

ConceptMapper


Taxonomic Object Service: SEEK

ConceptMapper

http://seek.nhm.ku.edu/TaxObjServ/services

TCSTCS

Find All Concepts

Get Synonymous Concepts

Get Best ConceptTOS

SEEK Cache

LSIDAuthority

Morpho

Data Analysis

EML Datasets

Identify species

EML(TCS)

Mark up datasets


Recap…Re-emphasised the problems with Taxonomic Names

not good identifiers for organisms problem extends to most areas

characters, countries, habitats, vegetation types, genes…..

Shown that Taxonomic concepts are better for referring to organisms, specimens, observations… but

Need better systems for resolving taxonomic names/concepts Which require better information


Provide better tools for usersTo help taxonomists create better quality data

Better access to reference/legacy data Explore differences/similarities in existing taxonomies To create relationships between concepts Improved data can be made available to the general biology

community for incorporating into bio-referenced databases.

To help end users understand and use the data and its limitations Biologists can use tools to understand the impact of using

particular data on their analysis


Conclusion Science is complex (and therefore split into specialisms)

Identify the overlaps/linkages in the different domains Need useful approximations of things to simplify linked domain Need to understand the approximations or linking points well

Support re-composition, linking or building on the components Science is inherently changing

Science is full of legacy data Today’s scientific research is tomorrow’s legacy data

Track the changes in the data know when components or links have changed

Provide long-term persistent storage Any published scientific discovery should store the data as evidence Data needs to be accurately annotated

Sufficient to repeat analyses to test hypotheses


AcknowledgementsColleagues on the SEEK project NSF and EPSRC fundinge-Science Centre funding Colleagues in TDWG

Thank You

Questions…

data integration issues in biodiversity research jessie kennedy shawn bowers, matthew jones, josh...

Documents

visual tools

new scientific discovery

complex slide

exchange of data

data integration issues

data sets ontologies

aimee stewart slide

science environment