Literature Mapping with PubAtlas extending PubMed with a BLASTing interface

1
Literature Mapping with PubAtlas -- extending
PubMed with a BLASTing interface
D Stott Parker1, WW Chu1, FW Sabb3, AW Toga2, RM
Bilder3 1UCLA Computer Science Dept, 2Laboratory
of Neuroimaging, 3Dept of Psychiatry
Biobehavioral Sciences
Hypothesis Web Project NIH RL1LM009833
2
PubAtlas Literature Map
PubAtlas is a PubMed BLAST-query service for
two term sets/lexica
result contingency table for all queries (X
AND Y) where X,Y are terms in the two lexica
www.pubatlas.org
3
Lexica

• PubAtlas Lexica
• term definition pairs
• Term Name PubMed Query
• optional hierarchical structure
• Lexicon as
• concept base
• ontology
• user-defined term hierarchy (personalized MeSH
  hierarchy)
• domain-specific query language

4
Concept BLASTing
Concept BLASTing seeks useful associations,
much like microarray analysis
Lexicon2 Y hierarchy
Lexicon1 X hierarchy
PubAtlas
Literature Map (X AND Y) association table
MEDLINE / PubMed as a bioscience association base
5
Previous Work -- as an example
Desirable extension features
semi-automated generation of a review paper --
but thorough and remaining up-to-date
graph "PubMed" TIAB AND ("graph"
TIAB OR "network" TIAB OR "diagram"
TIAB) visual "PubMed" TIAB AND
("visual" TIAB OR "visualizing" TIAB OR
"visualization" TIAB ) friendly
"PubMed" TIAB AND ("friendly" TIAB OR
"flexible" TIAB) better interface "PubMed"
TIAB AND ("interface" TIAB OR "interaction"
TIAB OR "query" TIAB) ) exploration
"PubMed" TIAB AND ("exploration" TIAB OR
"explore" TIAB OR "discovery" TIAB
) summarization "PubMed" TIAB AND
(summariz TIAB OR digest TIAB) map
"PubMed" TIAB AND ("mapping" TIAB OR
"map" TIAB OR "mapped" TIAB) extraction
"PubMed" TIAB AND (extract TIAB OR
identif TIAB) relevance "PubMed"
TIAB AND ("relevance" TIAB OR "ranking"
TIAB OR "ordering" TIAB) powerful
"PubMed" TIAB AND ("powerful" TIAB OR
"extended" TIAB OR "advanced" TIAB)
AliBaba "AliBaba" TIAB AND "PubMed"
TIAB Anne O'Tate "Anne O'Tate"
TIAB BioIE "BioIE"
TIAB ClusterMed "ClusterMed"
TIAB ConceptLink "ConceptLink"
TIAB GoPubMed "GoPubMed"
TIAB HubMed "HubMed"
TIAB PubFocus "PubFocus"
TIAB PubGene "PubGene"
TIAB PubMatrix "PubMatrix"
TIAB PubMed Assistant "PubMed Assistant"
TIAB PubNet "PubNet"
TIAB PubReMiner "PubReMiner"
TIAB Relemed "Relemed" TIAB SLIM
"Muin M" au AND "SLIM"
TIAB VisualNet "VisualNet" TIAB OR
"Visual Net" TIAB XplorMed "XplorMed"
TIAB
previous PubMed extensions
6
PubAtlas -- interesting aspects

• PubAtlas as a tool for concept BLASTing
• Moving towards shared, user-defined query/concept
  languages
• Visual literature search with concept maps /
  literature maps
• Building on familiar association mining metaphor
• Extending PubMed with temporal indexing / concept
  evolution
• Real uses semi-automated reviews, knowledge
  mgmt, ...
• Applications in Phenomics
• Phenotypes are often naturally represented as
  queries
• Promising applications in interdisciplinary
  collaboration

7
Knowledge Management
Who at UCLA works on Dopamine Receptors?
Many possibilities for interdisciplinary
collaboration
8
People as Concepts
Lori Altshuler Altshuler Lori FAU OR
Altshuler LL AU Stephen Marder Marder Stephen
FAU OR Marder SR AU Carrie Bearden
Bearden Carrie FAU OR Bearden CE AU Ty
Cannon Cannon Tyrone FAU OR Cannon
TD AU Michael Phelps Phelps Michael FAU
OR Phelps ME AU John Mazziotta Mazziotta
John FAU OR Mazziotta J AU Paul Thompson
Thompson Paul M FAU OR Thompson PM
AU Arthur Toga Toga Arthur FAU OR
Toga A AU Roger Woods Woods Roger
FAU OR Woods RP AU Bob Bilder
Bilder Robert FAU OR Bilder RM
AU Nelson Freimer Freimer Nelson FAU OR
Freimer N AU . . .
Map of publications in which people X, Y both
occur as authors
9
Exploring Associations over Time
10
Extending PubMed with Time
Historical map of interdisciplinary collaboration
at UCLA over 10 yrs
1998
2000
2002
2004
2006
2008
11
Deeper Exploration
Visualization and interaction along with standard
mining of association data
12
Larger Lexica
For term sets of size M, N, PubAtlas submits
MN PubMed queries This can scale to hundreds or
thousands of terms
13
Phenomic Vocabularies as Lexica
Diverse, complex phenotypes can be represented as
queries (predicates) -- denoting the set of all
relevant documents
PubMed / MEDLINE central phenomics database
14
Query Expansion -- for Phenotypes
"nback"

?
N-back
("nback" OR n-back OR "wisconsin card sorting"
OR "sternberg" OR "working memory capacity" OR
"stroop" OR "choice reaction time" OR "paced
auditory serial addition" OR "pasat" OR "digit
span" OR "delayed match to sample")

• Queries (like n-back test) can be expanded with
  terms related to their target concept (like
  working memory), using statistical models to
  identify better expansions.
• Expansion can improve precision and recall of
  queries that are being used as models of
  concepts/phenotypes

15
Summary

• PubAtlas as a tool for concept BLASTing
• Lexica are concept bases / user-defined query
  languages
• PubAtlas constructs concept maps / literature
  maps
• Extends PubMed with temporal indexing
• Multiple features for exploration, visualization
• Real uses semi-automated reviews, who is doing
  what, ...
• Many interesting directions for further work
• Applications in Phenomics
• Phenotypes are often naturally represented as
  queries
• Promising applications in interdisciplinary
  collaboration

16
Thank you!