Biomedical Informatics & Cancer

National Press Foundation Dec 11, 2012
Subha Madhavan, Ph.D. Director, Innovation Center for Biomedical Informatics Lombardi Cancer Center, Georgetown University


“Please note that some of the views presented are solely mine and do not necessarily reflect those of the Georgetown University.”

Era Of Discovery – And Better Treatments

Study Reshaping Ways of Treating Breast Cancer
By Gina Kolata September 23, 2012


Era Of Scientific Transformation
• First genome sequenced in 1995
– the bacteria H. influenzae with a genome of 1,830,140 letters

• 1st draft of human genome finished in 2001 (~ 3 billion letters) • Hundreds of higher-order genomes done or in progress. • The world‘s current sequencing capacity is estimated to be 13 quadrillion DNA bases a year requiring 3 or more exabytes of storage (approximately 1/100 of the entire world‘s data storage) • The NIH-funded 1000 Genomes Project deposited 200 terabytes of data into the GenBank archive during the project‘s first 6 months of operation, twice as much as had been deposited into all of GenBank for the entire 30 years preceding • Million Genome projects – not too far way!

Yet, in many cases we cannot answer patient’s individualized questions

• Will my cancer spread? • Do I need chemotherapy after surgery for my cancer type? • What are the side effects of chemotherapy for me? • Are there any new drugs targeted for my type of cancer? • What are my chances of survival?

Real Cost of Sequencing

Sboner et al. Genome Biology 2011, 12:125

Biomedical Informatics Defined
Biomedical informatics (BMI) is the interdisciplinary field that studies and pursues the effective uses of biomedical data, information, and knowledge for scientific inquiry, problem solving, and decision making, motivated by efforts to improve human health.
© AMIA 2011

Interdisciplinary Nature of Biomedical Informatics

© AMIA 2011

What Has Biomedical Informatics Delivered So Far?
• Human Genome Assembly (3 Billion Bases) • Disease Gene Discovery (Over 10K entries in OMIM) • Personalized Medicine (e.g. Her2) • Predicting Drug sensitivity & Response (e.g. Warfarin)

Three Examples of Use of Biomedical Informatics in Cancer Care & Research

1. Genomics As An Information Science

• National and International Efforts
– – – – – – – – – Broad‘s Genome Space MSKCC cBIO TCGA Internal Cancer Genomics Consortium Multiple Myeloma Portal Oncomine Harvard‘s i2b2 J&J‘s Transmart Georgetown Database of Cancer (G-DOC)

Integrated Environment: G-DOC Links To Many Data Silos
Pathway Studio
• Systems biology • Literature

•Visualization networks

• Molecule Visualization Ingenuity Analysis
•Pathways, GO, Variants

•Genome Visualization

RedCap Clinical Research
Disease Classification Tools


Over 330 registered users from six countries

~250 unique hits/month

G-DOC – Under The Hood

Personalize Colorectal Cancer Therapy
- To obtain a multidimensional molecular ―portrait‖ of cancer - To find most informative combination of the biomarkers correlated with clinical outcome - To test the biomarkers in the clinic

Molecular Signatures Of Stage II Colorectal Cancer Relapse

• ‗OMICS‘ Data
– – – –
– – – –

Gene Expression DNA Copy Number Metabolites miRNA data
Demographics Pathology Diet Family History

• Clinical

Gene Expression Markers Separate CRC Relapse Patients From Non-Relapse Cases

Patient Level DNA Changes
Chromosome 4

Impacted Biological Functions

Next Generation Sequencing Big Data Analytics On The Cloud
Genomic Data De-identified Clinical data



Genome Data

Amazon S3 Bucket

Amazon EC2 Compute Instances

G-DOC Web Portal

Clinical Data

Amazon Glacier Vault

MongoDB EC2 Instances

Oracle RDS Instance

2. Pharmacogenomics

One Size Fits All – Does not Work!
All patients with same diagnosis Treating with same drug(s)

Responders and Patients not Experiencing Severe Toxicity

Non-Responders and Patients Experiencing Severe Toxicity

Pharmacogenomics – Use?
• Identify populations that should receive lower or higher doses of a drug • Identify responder populations (e.g. ultrarapid metabolizers) • Identify high risk groups

Individualized Cancer Care: Cause & Treatment
Tumor Genetics “Somatic Mutations” Personal Genetics “Germline Mutations”

Personalized Rx Plan: Right Drug Right Dose Right Patient Right Time

Tumor Genetics Influences Drug Choice
• Lung Cancer
– EGFR→ Tarceva – Alk → Xalkori – Adeno → Alimta

• Breast Cancer
– Her2 → Herceptin – ER/PR → Tamoxifen

• Colon Cancer
– EGFR → Erbitux – MI / 18q deletions / DCC → FOLFOX

• Melanoma
– BRaf → Zelboraf

Many Cancers Are Molecularly Driven

On average 75% of patients don‘t respond to the cancer drug given to them

Source: Wall Street Journal, 2011

Drug Dosing May Vary By Ancestry

Average daily doses of 3.4 mg in Asians, 5.1 mg in whites, and 6.1 mg in blacks are representative of the literature for these ethnic groups

Johnson J A Circulation 2008;118:1383-1393 Reproduced from Dang et al,6 with permission from the Annals of Pharmacotherpy.

PGx Biomarkers In FDA Drug Labels


Genotype Tests Aid In Therapeutic Strategy

Help Clinicians Make The Right Drug Choices Clinical Factors
Environmental Factors

Genetic Factors

Clinical Decision Support (e.g. Bayesian Analysis)

Right Drug Right Dosage

Bayesian Network

• Background
– A Bayesian network is a probabilistic graphical model that represents a set of random variables and their conditional dependencies. – Generalizations of Bayesian networks that can represent and solve decision problems under uncertainty are called influence diagrams

• GeNIe software
– Developed at the Decision Systems Laboratory, University of Pittsburgh. – It can be used to implement graphical decisiontheoretic methods, such as Bayesian networks and influence diagrams

Clinical Decision Support Using Genotype Data

3. Mobile Health

• • • • • Need to enhance patient experience in the clinic Clipboard to iPAD Standardized data collection Error detection and data validation Effective secondary use of clinical care data – Collect once and use multiple times (for care, research, education) • Enable clinical research • Patients need to feel part of the process • Making patient education fun for all involved

Electronic Data Capture To Enhance Patient Engagement In Cancer Research & Care

Patient Navigation

Real-Time Clinical Decision Support
 Tonic‘s built-in risk assessment engine means that any patient form can also screen patients for a host of conditions that their physician may not have time to uncover, helping lead to improved –and more preventative—care, and the ability to better segment and manage patient populations….in real time.

Risk Assessments


(i.e. Autism, Anxiety, Alcoholism, etc.)

Triage Flagged Answers Clinical Trial / Research Eligibility

Real-time Clinical Decision Support (contd)
Example of web report. Can download, print or .pdf each patient report.

Example of printed pedigree report


Patient Education
 Tonic‘s real time data analytics engine also allows for patient education (such as videos, links or text) to be dynamically served up based on a patient‘s answers.
- Example: Tonic automatically screens patients for breast cancer risk and serves up a video about mammograms at the end of the questionnaire to those who are calculated as being at high risk. (Those at medium and low risk see different videos.)

 Risk scores can also be communicated to patients in real time, if desired.

Current Challenges
• Educating the next generation of physicians and researchers in information sciences • Clinical validation of research data • Data access and policy Issues • Varying data formats needing harmonization • Lack of clear, universal consent forms • Privacy issues around storing genomes on the cloud

My email: Our Center Website:

Sign up to vote on this title
UsefulNot useful