1.
Maria Shkrob, PhD, Project Manager, Elsevier Professional Services
m.shkrob@elsevier.com
May 19, 2016
Mobilizing informational resources
for rare diseases
When every piece matters

2.
| 2
Rare diseases – when every piece matters
Nick Sireau at TEDx ImperialCollege
https://www.youtube.com/watch?v=B4UnVlU5hAY
• No support
• No funding
• No treatments
is a UK charity that is building the rare disease community to raise awareness,
drive research and develop treatments.
is partnering with Findacure scientists to help identify and evaluate treatments
for congenital hypersinsulinism
• Patients community
• Collaboration with medical
researchers
• Drug repurposing candidate
• Fundraising
• Clinical Trial

3.
| 3
• A rare genetic disease
• Permanently excessive level of insulin
in the blood
• Develops within the first few days of life
• Can lead to brain injury or even death
• In the most severe cases the only viable
treatment is the removal of the pancreas,
consigning the patient to a lifetime of diabetes
• Sirolimus showed promising results in CHI
Congenital hyperinsulinsm
https://res.cloudinary.com/indiegogo-media-prod-
cld/image/upload/c_limit,w_620/v1440424745/uzvnqz
hvbpsrtthzxqpu.jpg
How can we help?

4.
| 4
Congenital hyperinsulinism library
In support of Findacure’s mission of education and knowledge sharing:
• Access to all Elsevier’s ScienceDirect full-text publications covering CHI
• Collection of papers focused on different aspects of CHI
• Collection of papers focused on effects of sirolimus on CHI

5.
| 5
Why do we need literature?
PLACES PEOPLE GENES
DRUGS INTERACTIONSPROPERTIES

6.
| 6
The power of processed content
PLACES PEOPLE GENES
DRUGS INTERACTIONSPROPERTIES
Data Extraction and Normalization
Databases and Tools

7.
| 7
• CHI Library
• Disease, Target, Pathway, and
Compound Analysis
• Research Landscape Analysis
Information Assets Applied
• Content
Elsevier’s vast set of literature and patent data
• Data normalization
Taxonomies and dictionaries to normalize
author names, institutions, drugs, targets, and
other important terms
• Information extraction
Finding semantic relationships, targets,
pathways, drugs, and bioactivities
Creating a comprehensive view of CHI with Elsevier
R&D Solutions

8.
| 8
Research landscape analysis: connecting patients,
researchers and institutions
0 10 20 30 40 50 60 70
Stanley, C.A.
Hussain, K.
De Lonlay, P.
Rahier, J.
Ellard, S.
Flanagan, S.E.
Shyng, S.L.
Nihoul-Fekete, C.
Bellanne-Chantelot, C.
Robert, J.J.
Brunelle, F.
KEY AUTHORS
0 10 20 30 40 50 60 70 80
The Children's Hospital of Philadelphia
UCL Institute of Child Health
Hopital Necker Enfants Malades
University of Pennsylvania, School of…
UCL
Universite Paris Descartes
University of Pennsylvania
Cliniques Universitaires Saint-Luc,…
University of Exeter
Oregon Health and Science University
KEY INSTITUTIONS0 1 2
Ajinomoto CO., INC.
Arkray, INC.
Korea Research Institute…
ViviaBiotech, S.L.
Bassa, Babu V.
Commisariat a l'Energie…
Glaser, Benjamin
Kowa CO., LTD.
Kyowa Hakko Kogyo…
KEY PATENTS
• Most prolific authors and institutions,
based on full-text searching for terms and
synonyms
• Patent assignee names from Reaxys

9.
| 9
Research landscape analysis: collaboration
• Network of people and organizations collaborating in CHI space based on
co-authorship

10.
| 10
High level summary of full text publications
Tag cloud of titles and sentences discussing hyperinsulinism:
• Provides a very high level summary of a group of publications
• Gives overview of the terms and words being used when discussing the
disease
Sized by inversed document frequency (IDF),
colored by term frequency (TDF)
Sized by relevance, colored by trend

11.
| 11
Why text mining?
Amorphous information Structured information
Image Source: http://www.thesocialleader.com/wp-content/uploads/2011/03/paper-piles.jpg
Text mining: analyzing text to extract information that is useful for particular purposes
Text
mining
• Hard to deal with
• Hard to deal with algorithmically
• Not scalable
• Search
• Visualize
• Network analysis
• Scalable
• Compressed
20km

12.
| 12
Elsevier Text Mining – Natural Language Processing
and deep taxonomy based indexing
~26M MedLine abstracts
~7M Elsevier and non-
Elsevier full texts
Grant applications
Dictionary
Taxonomy
Natural Language Processing
engine
MORE EFFECTIVE DOCUMENT SEARCH (CHI Library)
INFORMATION EXTRACTION (Summarization of Literature)

13.
| 13
CHI: finding relevant documents for CHI library
Dictionaries and taxonomies for:
• Proteins
• Small Molecules
• Diseases
• Clinical Parameters
• Organisms
• Biological Functions
• Anatomical Concepts
• Cell Lines
• Medical and Research Procedures
• External Factors
• Measurements
• Relations
Finding documents that mention CHI

14.
| 14
• CHI in abstract or title
• CHI subtypes
• By publication type
• By study type
(including MeSH terms)
CHI: finding relevant documents
Indicate what to query
Filter by study type
Specify distanceFinding documents that mention certain
aspects of CHI

15.
| 15
CHI: finding relevant documents
TERM1 VERB TERM2
Target ----------------- Disease
Small Molecule ----- Target
Small Molecule ----- Disease
Disease -------------- Biomarker
Protein --------------- Process
Output literature that discusses
the relation of interest
Finding documents that mention effects of
sirolimus on insulin sensitivity, production
and release

16.
| 16
CHI: finding targets, drugs, and drug effects
"protein"
"terms for
genetic
variations"
"Persistent
Hyperinsulinemia
Hypoglycemia of Infancy"
Relevant Text Title Authors
Reference
Date
DOI
ABCC8 mutation Persistent
Hyperinsulinemia
Hypoglycemia of Infancy
In the literature, nine genes have been reported to
be associated with CHI , with the most common
genetic causes of CHI being mutations in either
ABCC8 or KCNJ11 .
Successful treatment of a newborn
with congenital hyperinsulinism having
a novel heterozygous mutation in the
ABCC8 gene using subtotal
pancreatectomy
Yen C.-F, Huang C.-Y,
Chan C.-I, Hsu C.-H, Wang
N.-L, Wang T.-Y, Lin C.-L,
Ting W.-H.
2016 10.1016/j.
tcmj.2016
.04.001
ABCC8 loss of function
mutation
Persistent
Hyperinsulinemia
Hypoglycemia of Infancy
GOF and loss-of function mutations in KCNJ11
(Kir6.2) and ABCC8 (SUR1), which encode the
predominant KATP channel subunits in
pancreatic β-cells and in neurons, are now well-
understood to underlie neonatal diabetes and
congenital hyperinsulinism, respectively.
Adenosine Triphosphate-Sensitive
Potassium Currents in Heart Disease
and Cardioprotection
Nichols C.G. 2016 10.1016/j.
ccep.201
6.01.005
ATP-activated inward
rectifier potassium
channel
mutation Persistent
Hyperinsulinemia
Hypoglycemia of Infancy
The prevalence of KATP channel gene mutations,
diazoxide responsiveness, and rates for surgery
is broadly commensurate with other CHI cohorts.
Feeding Problems Are Persistent in
Children with Severe Congenital
Hyperinsulinism
Banerjee I, Forsythe L,
Skae M, Avatapalle HB,
Rigby L, Bowden LE,
Craigie R, Padidela R,
Ehtisham S, Patel L,
Cosgrove KE, Dunne MJ,
Clayton PE.
2016 10.3389/f
endo.201
6.00008
Extracting structured information from text
Standardized
names
Standardized
link
Evidence

17.
| 17
CHI: summarization and visualization of the findings
• Visualization and summarization of
6.2 M literature findings
• Linking to non-literature sources

18.
| 18
Building and refining the disease model
Picked relevant
pathways
(from a collection of 1800 models)
Explored functions of
proteins using 6.2M pre-
text mined relations
and embedded Gene
Ontology
Summarized what is known
about CHI mechanism in an
overview model

19.
| 19
CHI: Building and refining the disease model

20.
| 20
From pathways to treatments:
PipelinePilot implementation combines data sources
Automated analysis combines bioassay data with text-mined data
Find all targets that could
be used to affect the
disease state
Query for each protein to find
compounds that target it (>6
log units)
Collate data by compound to summarize the
targets/activities related to disease that the
compound hits
• Compute geometric mean of activities for ranking
• Rank by number of targets and geometric mean of
activities against targets
Step 1 Step 2
Step 3

21.
| 21
Automated analysis combines bioassay data with text-mined data
From pathways to treatments
• 88 targets related to
hyperinsulinism with ≥3
literature references
• Full relationship
information
Find all targets that could
be used to affect the
disease state
Step 1

22.
| 22
Automated analysis combines bioassay data with text-mined data
From pathways to treatments:
Find all targets that could
be used to affect the
disease state
Query for each protein to find
compounds that target it (>6
log units)
Step 1 Step 2
Targets based on
text mining
Approved
compounds
Bioassay data

23.
| 23
Automated analysis combines bioassay data with text-mined data
From pathways to treatments:
Mean of activities
among these targets
Mean of activities
among these targets
Targets and activities
for each compound
Drug-likeness
metrics for
sorting/classification
• All compounds that
were observed to bind
to targets in pathway
• Sorted by number of
active targets.
Too many targets may
suggest lack of specificity.
Find all targets that could
be used to affect the
disease state
Query for each protein to find
compounds that target it (>6
log units)
Collate data by compound to summarize the
targets/activities related to disease that the
compound hits
• Compute geometric mean of activities for ranking
• Rank by number of targets and geometric mean of
activities against targets
Step 1 Step 2
Step 3

24.
| 24
Approved compounds that may treat hyperinsulinism
• Each binds to one or
more targets related to
the disease
• Can easily be obtained
and tested in preclinical
studies
• List includes a
compound known to
treat hyperinsulinism,
sirolimus

25.
| 25
From pathways to treatments:
PipelinePilot implementation output
Input:
“Congenital hyperinsulinism”
Output:
• Table of target information
(PathwayStudio)
• Table of compounds with targets,
activities, and druglike parameters for
each compound
• SD file of compounds that may be
efficacious, with clinical status
• Authors, Affiliations, Collaboration map
• List of papers

26.
| 26
Power of combining
pathway data with
experimentally verified
binding data
Results in testable
ideas
• Many compounds are
already approved drugs,
can be tested in in-vivo
experiments
Concepts can be extended
to find novel compounds
• Use modeling tools to extract
common frameworks
• SAR to optimize activity for
new indication
• Compare with compounds
suggested as treatments as
found by text mining
From pathways to treatments:
PipelinePilot implementation summary

27.
| 27
Findacure: empowering patient groups and facilitating
treatment development
Parents:
• Learn more about the disease
• Find doctors and medical centers
Doctors:
• Learn more about the disease
• Explore case studies
• Collaborate
Researchers:
• Testable ideas for repurposing of generic drugs
• Knowledgebase to support the research of the disease
mechanisms
• Collaborate
Evidence to support 10 drug repurposing trials

28.
| 28
• Used extensive Elsevier’s content, tools and capabilities to provide
information about a rare disease:
 Text Mining to find targets and summarize what is known about the
disease mechanism
 Bioactivity data to find drugs that target those targets
 Normalized names of authors and institution to find collaborators
• Once the output of interest is decided, answer generation can be
automated:
Provide disease name and get:
 List of targets with supporting information
 Sorted list of approved drugs with supporting information
 KOLs and institutes
Summary

29.
| 29
Findacure / Elsevier collaboration
Dr Rick Thompson
Findacure
Dr Nicolas Sireau
Findacure
Dr Matthew Clark
Elsevier
Dr Maria Shkrob
Elsevier

30.
Thank you
https://www.elsevier.com/solutions/professional-services