Transcript Xiaoqian Jiang, PhD

MED 264 introduction
Xiaoqian Jiang, PhD
The analyses upon which this publication is based were performed under Contract Number HHSM-500-2009-00046C sponsored
by the Center for Medicare and Medicaid Services, Department of Health and Human Services.
MED 264 Introduction
• Introduction and class overview
• Topics and expectations
• A brief introduction of biomedical informatics
2
MED 264
• 22 students
• 10 weeks, 2 course per week
• Class website: http://course.ucsd-dbmi.org/MED264/
• [email protected]
3
Date
10/2/2014
10/21/2014
Lecturer
Title
Xiaoqian Jiang Introduction to MED264
Mary Linn
Systematic Reviews: principles and processes
Bergstrom
Public health information systems and interoperability and data standards in public health
Mike Hogarth
informatics
Zhuowen Tu Introduction to information retrieval and data fusion
Lucila OhnoResearch methods (study design, sample size, evaluation of models)
Machado
Claudiu Farcas Project management and software engineering related to informatics projects
10/23/2014
10/28/2014
10/30/2014
11/4/2014
11/6/2014
Shuang Wang
N/A
Yunan Chen
Jihoon Kim
Cui Tao
Introduction to R with Shiny for sharing interactive biomedical research results
No lecture scheduled (due to conference)
Evaluation of information systems to provide feedback for system improvement
Statistics for biomedical research
Applying Ontology and Semantic Web Technologies to Clinical and Biomedical Studies
11/13/2014
11/18/2014
11/20/2014
Edna Shenvi
Chun-Nan Hsu
Son Doan
Robert ElKareh
Xiaoqian Jiang
Cleo Maehara
Impact of clinical information systems on users and patients
NLP applications in biomedicine
Introduction to biomedical natural language processing
10/7/2014
10/9/2014
10/14/2014
10/16/2014
11/25/2014
12/2/2014
12/4/2014
12/9,11/2014
Clinical Decision Support
Privacy policy and technologies for healthcare research
Imaging informatics
4
Presentations
Grading Policies:
• Course grades will be based on
1) Attendance (10%),
2) Assignment and mid-term project review (30%),
3) Project oral presentation (15%),
4) Project participation (15%),
5) Final project report (30%)
5
Biomedical Informatics
Algorithms
Controlled vocabularies
Ontologies
Data management
Information retrieval
Pharmacogenomics
Personalized
Medicine
Electronic Health Records
Decision Support Systems
Hospital Information Systems
Healthcare
Systems
Medical
Informatics
Bioinformatics
Genomics
Transcriptomics
Proteomics
Epigenetics
Big Data
Today:
• Some data on a lot of individuals
– Example: observational data from EHRs
• A lot of data on some individuals
– Example: sensor data
Tomorrow:
• A lot of data on a lot of individuals
– International collaborations
Personalized Care and Population Health
• Genomics
– SNP-based therapy (cancer)
• ‘Phenomics’
– Electronic Health Records
– Personal monitoring
• Blood pressure, glucose
– Behavior
• Adherence to medication, exercise
• Public Health and Environment
– Air quality, food
– Surveillance
Source: DOE
UC ReX - Research eXchange
• Clinical Data Warehouses from 5 Medical
Centers and affiliated institutions (>10
million patients)
• Aggregate and individual-level patient
data to be exchanged according to data
use agreements, internal review boards
• Funded by the University of California
Office of the President
9
iDASH
10
Integrating Different Types of Data
genome
transcription
RNA
transcriptome
translation
Genotype
Protein
Phenotype
physical exam, imaging,
monitoring systems
Physiology
tests
Metabolites
proteome
laboratory
What can we do?
• Build access to large data repositories to
improve research
– Enhance policy and technological solutions to the
problem of individual and institutional privacy
– Donate data
• Aggregate data from different countries and use
for new analyses
– Provide tools to integrate and analyze data
Privacy Protection
– Use of clinical, experimental, and genetic data for research
• not primarily for clinical practice (i.e., not for health care)
• not primarily for quality improvement (i.e., not for IRB exempt
activities – regulatory ethics committee)
– iDASH will host and disseminatte data according to
• Consents from individuals
• Data owner (institutional) requirements
• Federal and state rules and regulations
funded by NIH U54HL108460
13
Shared Model Building and Evaluation
Wu Y, Jiang X, Kim J, Ohno-Machado L. Grid Binary LOgistic REgression (GLORE): Building
Shared Models Without Sharing Data. JAMIA 2012
14
Personalizing Medicine
Prevention
– Risk Assessment
• Genomics
Diagnosis and Therapy
– Decision support
• Pharmacogenomics
Big Data
- Secure Cloud Environment
• Electronic Medical Records
• Genetic Data
22%
16%
Your Risk
“this program
shows the
estimated
health risks of
people with
your same
age, gender,
and risk factor
levels”
p=1
x
“this means that 5 of 100 people
with this level of risk will have a
heart attack or die”
People “like you”
Input space
Output space
p
“people with
your same
age, gender,
and risk factor
levels”
p=1
x
“people with this
level of risk”
Who should get a liver transplant?
14
10
20
14
p
risk
Individualized Confidence Interval
Large Individual Confidence Interval
Probability
estimate
Narrow C.I.
25
Patients “like you” get predictions like you, but
different confidence intervals
height
Probability Estimate = 0.3
C.I. = [0.05, 0.55]
14
risk
20
Probability Estimate = 0.3
C.I. = [0.2, 0.4]
10
1
1
me
14
me
p
gender
Confidence Interval (CI) Near the Boundary
27
Far from the Boundary
28
C.I. depends on Density
2011 summer
internship
program
29
Sparse region, larger C.I.
30
Adaptive Calibration
Large Individual Confidence Interval
Probability
estimate
Narrow C.I.
31
Adaptive Calibration
Recalibrated prediction 2/4 = 0.5
Probabilit
y
estimate
Recalibrated prediction 1/3 =0.33
Jiang X, Osl M, Kim J, Ohno-Machado L. Calibration of Predictive Model Estimated to Support Personalized
Medicine. J Amer Med Inform Assoc 2012
32
Adaptive Calibration of Predictions
33
Original Estimates
34
Recalibrated Estimates
35
Who should get a liver transplant?
ELIGIBLE FOR TRANSPLANTATION
1
2
0
1
1
me
1
me
p
risk
NOT ELIGIBLE FOR
TRANSPLANTATION
Biomedical Informatics
• Data compression
• Dimensionality
reduction
• Information retrieval
• Data annotation
• Visualization
• Genotype-phenotype
associations
• Temporal associations
Research Education
Service
Change