More Related Content Similar to SciBite Short Intro Sept 2015 (20) SciBite Short Intro Sept 20151. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
Termite
Text Analytics
September 2015
SciBite: Turning Text Into Intelligence
2. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Termite: Text Analysis Engine
Identifies the
genes,
diseases,
drugs, devices,
etc in
biomedical
text
3. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Termite: Advantages
SS
Quality
High curated
vocabularies, many fold
enriched over those in
public domain
SS
Ambiguity
“Hedgehog”, “GSK”,
“ACE”,
“pdf (=peptide
deformylase)”
SS
Coverage
Beyond the normal “Gene,
Disease, Drug” to include
PK/PD, ADME, Med.
Devices, Chemistry
Methods, Pathogens,
Business & more
SS
Speed
Analyses up to 1 million
words/second
Text-mine the entire
Medline database on
a laptop in a few
hours
SS
Its Live!
Simple API –send it a
document, it sends
you back results …
immediately
SS
Usability
Modern, Java-based REST-
ful service
Easy to set up. Runs on a
server, cloud, laptop
Batch Mode, Pipeline
Pilot, Your Code
4. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Embedding
Text Analytics
In Applications
5. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Embedding Termite In Apps - Why?
A Better Search
Experience
Find documents that mention “Lipitor” when the
search term is “Atorvastatin”
Concept-Type Searches Find any documents that mention a gene or
indication and my topic of interest
Summary Perspectives Ask “What are all the targets discussed in documents
concerning topic such as drug repurposing or a
particular drug or indication”
Ontology Queries Find any documents that mention a kinase or
inflammatory disorder
Transformative Data
Integration
Add structure to unstructured data in documents and
connect it to databases and other systems to provide
a complete view across the organisation
6. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Termite + Sinequa (http://sinequa.com)
Rich,
semantics-
based
facets
7. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
http://news.scibite.com
If genes could tweet….
8. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Enhanced Medline & Correlation Networks
9. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Live Integration:
SureChEMBL UI & SciNav
Ack. ChEMBL Group,
particularly Nathan
Dedman
10. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Termite UI Kit
Embedding into
ELNs, Registration Systems,
Project Dbs, Sharepoint etc
Semantic Autocomplete
11. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Summary
• Many more use-cases for “semantic
enrichment” within apps
• Better User Experience
• Better Storage Of What Really Matters
• Facilitates Analysis & Links
12. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Data Analyst
Use Cases
13. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Semantic Regular Expressions
(Singulair/Montelukast)
Find me all the reported biomedical effects of the drug, Singulair
14. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
{INFL#INF3}induced {PROTYP#PROTYP702}cc
_chemokine
{$SCIVERB#PRODUCTION}production leading
to {INDICATION#D001249}asthma
{INFL#INF6}inhibition of {GENE#CCL2}ccl2 with
neutralizing antibody significantly
{INFL#INF8}attenuated hrv {INFL#INF3}induced
{INDICATION#D001249}airways _inflammation
{INFL#INF6}inhibition of
{PROTYP#PROTYP576}poly _adp _ribose
_polymerase {INFL#INF9}prevents allergen
{INFL#INF3}induced
{INDICATION#D001249}asthma
{INFL#INF6}inhibition of {GENE#TNF}tumour
_necrosis _factor _alpha may be useful in
severe {INDICATION#D001249}asthma
Example: Disease Modifying Genes
Problem: I want to know all
genes that have a modifying
influence on my input condition
(e.g. asthma).
Solution:
• Create a pattern for gene ~
influencing verb ~ disease
• Run on document subset
15. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Biomarker Discovery
Obtain Breast Cancer Articles
From Medline
Filter By Body Fluid
Analyse Gene-Disease
Relationships Using
1. Abstract CC
2. Sentence CC
3. “is a marker of” Pattern
Found All + More!
16. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Termite On EHRs
• Example protocol “Tell me the most common drugs taken by patients in the
following set of Electronic Health Records”
• Uses a sample EHR set from the Translational Medicine Ontology Project,
Luciano et al, J Biomed Semantics. 2011; 2(Suppl 2): S1
17. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Which
Patients
Where Mentioned In
EHR
Donepezil most
common in 7
patients
Followed by
Carbamazepine
in 5 patients
..
etc
18. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Find New Terms (Term Extractor Workflow)
We want to identify text terms that
were not recognised by Termite, but
look significant
• Fetch articles with “herbicide” in
the title
• Scan with Termite, identify non-
entity, but significant terms
• Group, count, rank
19. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Targets In Patents
! Over 50 targets mentioned !
SRC proto-oncogene, non-receptor tyrosine kinase ; aurora kinase A ; insulin ; microtubule-associated protein tau
; catenin (cadherin-associated protein), beta 1, 88kDa ; interleukin 1, alpha ; c-src tyrosine kinase ; phosducin-like
2 ; colony stimulating factor 2 (granulocyte-macrophage) ; tumor necrosis factor ; glycogen synthase kinase 3 beta
; amyloid beta (A4) precursor protein ; asparaginase ; interferon, beta 1, fibroblast ; jun proto-oncogene ;
coagulation factor II (thrombin) ; acetylcholinesterase ; FGR proto-oncogene, Src family tyrosine kinase ; BLK
proto-oncogene, Src family tyrosine kinase ; angiotensin I converting enzyme ; albumin ; axin 1 ; heat shock
transcription factor 1 ; v-myb avian myeloblastosis viral oncogene homolog ; NADH dehydrogenase, subunit 1
(complex I) ; LYN proto-oncogene, Src family tyrosine kinase ; LCK proto-oncogene, Src family tyrosine kinase ;
CCAAT/enhancer binding protein (C/EBP), alpha ; HCK proto-oncogene, Src family tyrosine kinase ; ATP citrate
lyase ; Glycogen synthase kinase 3 ; beta-catenin ; interferon ; tnf superfamily ; lactate dehydrogenase ;
neurotrophic factor ; pyruvate kinase ; fibroblast growth factor ; glycogen synthase ; tumor necrosis factor alpha ;
serine threonine kinase ; tau protein ; interleukin ; albumin ; thrombin ; eif2b ; ion channel ; cytokine ; heat
shock factor ; axin ; cAMP Response element binding protein ; amyloid beta ; Microtubule Interacting Protein ; c-
Jun N-terminal kinases ; src homology ; calcium channel ; atpase ; acetylcholinesterase
20. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Patent Relevancy Workflow
The method of claim 27, wherein the
method comprises inhibiting Aurora-2,
GSK-3, or Src activity
21. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
SciBite
Conclusion
Simple, Flexible Integration Through Semantics
For Data Analysts AND Applications
22. © 2015 SciBite. Registered In England & Wales No. 07778456 http://scibite.com | @Scibite | info@scibite.com
Phil Verdemato
Robert Greenwood
Dave Burrows
Ian Harrow
ChEMBL
Open PHACTS
Thanks & Acknowledgements