Transcript New Relationships Medline Full (prev. 5 yr)

Innovative Activities @ Novartis Knowledge Center
-- Create and deliver value beyond traditional information delivery
Richard Cai
Competitive Intelligence Manager
Novartis Knowledge Center
Philadelphia, June 2016
Topics
Innovations to add value beyond traditional information
delivery at Novartis Knowledge Center
• Innovative Alert Sharing
• Text Mining (highlighted later)
• Expert Identification
• Social Network Analysis
• ...
2 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
Novartis Knowledge Center (NKC) Overview
 NKC is THE information service provider for Novartis
• Who we support
• What we do
- General support
• Resource/vendor management
• Document delivery
• Training
- Special services
3 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
New environment, new roles
Enabling scientists to reach their full potential by
Visualization
allowing them to manage and interpret more of the
information made readily available to them
Advising individuals and teams on the best
Analysis
Information
Consultant
Search
resources for their current needs and enabling
them to get the most from those resources
Consulting clients who have special requests
with detailed analysis and insights
Sources
Traditional Search (AND/OR/NEAR/...)
4 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
Going beyond traditional information delivery @
NKC
AlertMe
Expert ID /
Doc Tagging
5 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
Text Mining
NetWorking
Current and Future View
 Expertise Location
 Network Analysis
 Community/Social Media
 Automated Workflow
 User Customization
 Integration of Tools
/Outputs
6 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
Goals and Strategies of a Pilot
-- can we facilitate compound repurposing
All NVS Past &
Present Targets
(regardless of
indications)
Medline
Full
(prev. 5
yr)
with NVS targets in
MESH
Genes
related to
disease
Linguamatics Search:
NVS Targets – relationship – Retinal Diseases
Medline,
Weekly
Updates
Retinal
Diseases
New Relationships
without gene match in
MESH, with long disease
names
Genes
related to
disease
7 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
Calculate
occurrences
w/in the prev.
5 yrs
Doc
Genes ranked
by
occurrences
Statistics
Total hits
“New” rel. hits*
Tech. wrong hits**
1241
145
63
# “New” rel. hits
# Wrong hits
Among wrong hits:
 All with misidentified
genes
 Four also with
misidentified diseases
Average around 4.5 hits/wk
wk



Users normally go directly to abstracts, skipping the gene
Users estimate that 25% of the time they find the papers interesting enough to
download the full articles
Wrong genes not necessarily lead to non-relevant papers. This is probably because
we do a good job getting the diseases right.
*: “New” rel. hits: hits that reveal relationships that have not been published or only once in the past 5 years.
**: Wrong hits: hits with misidentified genes or diseases. These are judged technically, not scientifically.
8 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
Some Thoughts
• Can search with large
#s of terms, including
synonyms
• Can customize complex
query structures
• Works well with
structured data
9 | Presentation Title | Presenter Name | Date | Subject | Business Use Only
• Not a general tool
• Requires expertise
and has a relatively
steep learning curve
• Depends on
Linguamatics for
indexing
More Thoughts
 Quality of the dictionaries
 Flexibility of the tools
 Get it right (from pure technology perspective) vs Get it
right (from customer’s need perspective)
10 | Presentation Title | Presenter Name | Date | Subject | Business Use Only