Transcript Watson Genomics Advisor

Watson Genomic Analytics
Select Watson solutions address a wide range of clinical and
research needs in oncology
Patient Insights
Electronic Medical Record
Advisor
Evidence-based Insights
Research Insights
Watson for Oncology
(Lung, Breast, Colon/Rectal
Treatment Plans)
Watson Clinical Trial Matching
(Identify all eligible trials for a patient)
Watson Discovery Advisor
(Insights from vast Medical and Research literature)
Watson Genomics Advisor
(Insights into Tumor DNA Sequencing)
Analysis of Medical Images (MRI, Mammogram, etc)
Available today
© 2014 International Business Machines Corporation
Currently in
Development/Testing
Research Phase
2
Path Toward Personalized Medicine
Green, ED et al (2011). Charting a course for genomic medicine from base pairs to bedside. Nature 470: 204-213
Change in personalized
healthcare investment
from 2005 to 2010 1
Biopharmaceutical companies
investing in personalized
healthcare research in 2010 1
75%
94%
1
Prominent personalized
medicine treatments &
diagnostics available 2
13
113
in 2006
in 2014
Tufts Center for the Study of Drug Development, 2010; 2 Personalized Medicine Coalition, 2014
© 2014 IBM Corporation
Decreasing Cost of Genome Sequencing
© 2014 IBM Corporation
Cancer Workflow: Research and Patient-Care
Genes
Cells
Proteins
Organs
Tissues
Organ
Systems
Body
Mutation
Dysfunction
Hyperplasia
Mass
Basic
Science
Dysfunction
Symptom/Finding
Medical Hx
Family Hx
Physical Exam & Review of Systems
Primary
Care
•
•
Radiology
Diagnosis
Sub-type Analysis
Biopsy
Histology/Cytology
Tumor Markers
Molecular Analysis
Chemotherapy
Oncologist
•
•
Personalize Therapy
Apply Treatment
Guidelines
Radiation
Surgery
Biologics
© 2014 IBM Corporation
Diving Deeper on Gene to Protein Relationship
© 2014 IBM Corporation
Survival Benefit of Targeted Treatment
Kris M, et al. Lung Cancer Mutation Consortium Survival by Group 2014, American Medical Association.
© 2014 IBM Corporation
How are These System Being Developed?
Clinical trials
Clinical
Treatments
Ingest
Patient Reports
Reference
Genomes
Chemical
Mutation
Learn
Pharmaceutical
Reports
Protein
Pathways
Genomic Data
Scientific
Literature
Patents
Medline
Test
Dysfunctional Proteins &Targeted Treatments
© 2014 IBM Corporation
Protein Pathways: Consensus
© 2014 IBM Corporation
Protein Pathways: Exploratory
Natural Language Processing (Annotators) Identify and Provide Structure to Concepts
…doxorubicin results in extracellular signal-regulated kinase (ERK)2 activation,
which in turn phosphorylates p53 on a previously uncharacterized site, Thr55…
Ingest
Learn
Test
ERK2
phosphorylates
p53
on
Thr55
Extracts Entities
 ERK2 = Protein, P53 = Protein, Thr55 = Amino Acid
Extracts Verb
 Maps to domain of Post Translational Modification
 Recognizes subject / object relationships
Extracts Entities
 ERK2 = Protein, P53 = Protein, Thr55 = Amino Acid
Extracts Preposition
 Recognizes preposition location on Thr55
Extracts Entities
 ERK2 = Protein, P53 = Protein, Thr55 = Amino Acid
© 2014 IBM Corporation
Protein Pathways: Exploratory
Concepts are Classified and Relationships Defined
Annotator
Logic
Ingest
Learn
Test
• Drug = entity
• Side effect = entity
association cause
• Cause = relating verb
• Rule = 1 drug to 1
side effect
Apply Annotators to
Text
• Aspirin is an antiplatelet indicated to
reduce the risk of myocardial
infarction
• Known side effects include
Gastrointestinal (GI) pain, GI upset,
ulcers, GI bleeding, and nausea
• Valium or Diazepam is a
benzodiazepine derivative, indicated
for the treatment of anxiety, muscle
spasms
• Valium may cause depression,
suicidal ideation, hyperactivity,
agitation, aggression, hostility…
Watson Creates
Knowledge Graph
GI Pain
Aspirin
Valium
Depression
© 2014 IBM Corporation
Protein Pathways: Exploratory
Knowledge is Reviewed and Statistics Added
Question
Ingest
Learn
What genes
contribute to
developing
colon cancer?
Search
Corpus
Score & Weigh
• Side Effects
• Quantity
• Lab Notes
• Proximity
• Genes
• Relationship
• Publications
• Domain Truths/
Business Rules
• Drugs
Test
Extract
Evidence
• Animal Models
• Clinical Trial
Data
© 2014 IBM Corporation
© 2014 IBM Corporation
Protein Pathways: Exploratory
Overall
FUNCTION
Step 4: Prediction
Known Pathways
Jak3
Jak1
Predicted Effects
P53
Jak2
TCF5
TCF7
ATM
SER1
Step 3: Relationships
Gene A
FORM
or
Gene B
Ontologies
Step 2: Organization
(e.g. organism, cell, protein,
amino acid)
Step 1: Exploring for Entities
Unstructured
Domain Entities
© 2014 IBM Corporation
Exploring Scientific Literature
© 2014 IBM Corporation
Exploring Scientific Literature
© 2014 IBM Corporation
Watson Genomic Analytics
© 2014 IBM Corporation
What Genomic Data is Being Leveraged?
Sample Collection
Variation Detection
Sequencing
Data Presentation
© 2014 IBM Corporation
Watson Genomic Analytics: Process
 Molecular Profile Analysis
Input: (Patient Specific)
1) Somatic Mutation (VCF or MAF file)
2) Copy Number Variation (log2 format)
These patient specific abnormalities are
compared against known mutations and
reference genomes to determine likely
“drivers” of the patients cancer
-Databases are gathered from consensus
community leading
Output: (Clinically Focused)
1) List of Dysfunctional Proteins
2) IBM Developed Driver Score
3) Targeted Therapies
© 2014 IBM Corporation
Watson Genomic Analytics: Process (continued)
 Pathway Analysis
 Drug Recommendation
Collections of consensus pathways (known)
and NLP based augmented pathway
(unknown) is used for our pathway traversal
algorithm
Proteins directly or closely related mutated
proteins are identified and correlated with
approved or investigational drug therapies
© 2014 IBM Corporation
Connecting Mutations to Treatable Targets
© 2014 IBM Corporation
Summary
 As sequencing becomes less resource intensive genomic data is becoming more and more
prevalent.
 Genomic Data is being integrated with scientific literature and patient data to advance
clinical care. This integration is allowing personalized medicine to take shape.
 In response to the continued growth in the amount and complexity of medical knowledge
industry leaders are leveraging process and machine-learning algorithms to scale
expertise within and across the various basic science and clinical domains.
© 2014 IBM Corporation
© 2014 IBM Corporation