platform for personalized oncology

7/28/2019 Platform for Personalized Oncology

1/33

Platform for Personalized

OncologyIntegrative analyses reveal novel molecular signatures

associated with colorectal cancer relapse

Subha Madhavan, Ph.D.

Innovation Center for Biomedical InformaticsGeorgetown University Medical Center

AMIA TBI, 2013


2/33

Driving Factors

Need for an information continuum (care -> research

-> back to care)

Incorporate omics-based evidence in clinical

research and care settings

Collect data once and use it multiple times quality

control, secondary use for research

Connect research platforms to accelerate scientific

discovery

Maintain data provenance and key findings to support

funding opportunities


3/33

G-DOC links to many information silos

G-DOC

Pathway Studio Systems biologyanalysis

Literature mining

Ingenuity VariantAnalysis

Variants Pathways, GO,Literature

Clinical Research

REDCap

Medidata

EHR

ARIA

AMALGA

CENTRICITY

JMol/Marvin

3-D Structure and

Molecule Visualization

Cytoscape

Visualization networks(pathways,interactions)

JBrowse

Genome Visualization

Heatmap Viewer

Visualization of copynumber data


4/33

G-DOC information flow


5/33

https://gdoc.georgetown.edu

Over 350 registered

users from six countries

~250 unique hits/month


6/33


7/33


8/33


9/33


10/33

Correlate abnormality/event with clinical

parameters


11/33

Problem:

Colorectal cancer (CRC) is the third most commonly diagnosed cancerin the United States in both men and women

In 2013, an estimated 143,460 new cases are expected to be diagnosed,

and 51,690 deaths from CRC are expected to occur in the United States

Approximately 75- 80% of patients with stage II colon cancers will be cured

by surgery alone, however, 20-25% will still experience relapse andultimately die of metastatic disease.

Goals:

To obtain a multidimensional molecular portrait of cancer relapse

To find most informative features correlated with clinical outcome

To identify novel combinations of different types of biomarkers of relapse

To use the data for uploading to G-DOC web portal

as a pilot multi-omics profiling study for further in silico analysis

What can we do with this platform?

Colorectal Cancer Multi-Omics Study


12/33

Two workflows for identification of major players

Biology-driven approach:

Systems Biology analysis of differentially expressed

features i.e. genes, microRNA etc

Results: Major pathways and Bio functions affected

in patients with relapse Data-driven computational approach:

Machine Learning algorithms: SVM, RF-ACE

Results: Small number of most informative features

associated with Relapse

Question: would these results converge on the same biology?


13/33

Gene Expression Analysis


14/33

Genes in tumor samples: Relapse vs. Relapse free

PCA based on T-test p


15/33

Enrichment Analysis: Diff. Expressed Genes

Bio-Functions Most Affected In Relapse Group


16/33

Sub-network enrichment: Cell processes

Gene Set Seed Overlapping Entities p-valueTotal # of

Neighbors Overlap

inflammatory response

EGFR,VCAM1,TNFSF11,IGF2,MAPK3,TACR3,CD5,NOS2,CSF3,HBEGF,HSPB1,CASP1,

APP,RETN,ILK,HSPD1,EGR1,TACR1,CXCL10,CXCL11,CXCR2,SELL,PPARD,WNT5A,IR

F1,ALOX15,CXCL1,PYCARD,SCARB1,IL1RN,IL7R,MMP7,TLR9,STAT1,LY96,TACR2,PRSS1,FPR2,ATF4,MMP12,FCGR3A,CFH,ICOS,CTLA4,IDO1,VIPR1,CD86,CXCL9,C4BPA,

MC3R,GPR44,IL13RA2,CHI3L1,CXCL3,LCP2,SERPIND1,CD74,NLRP2,APOL1,GC,IRAK

3,CLEC4E,BTN1A1,CD300C,TNIP3,DST,IL18R1,FREM1 8.13E-12 1237 68

immune response

TRIM56,EGFR,VCAM1,TNFSF11,MAPK3,TG,CD5,NOS2,CSF3,HSPB1,CASP1,APP,RET

N,HSPD1,EGR1,TACR1,CXCL10,PTPRC,CXCL11,CXCR2,SELL,PPARD,IRF1,CXCL1,P

YCARD,IL1RN,IL7R,MMP7,TLR9,STAT1,LY96,TXNIP,TACR2,BCL2A1,FPR2,VEGFC,FC

GR3A,CFH,ICOS,KLRK1,CTLA4,IDO1,ORM1,VIPR1,CD86,CXCL9,C4BPA,RCAN1,GPR4

4,TNFRSF13B,CXCL13,CRY2,F13A1,IL13RA2,CHI3L1,IBSP,PSMB9,LCP2,PIGA,ADAM8,

GZMM,CR2,TAP1,ERVWE1,CD74,LBR,LAMP3,CSTA,UBD,APOL1,PYHIN1,GC,IRAK3,C

LEC4E,FYB,BTN1A1,CD300C,S100A13,CLEC1A,MOAP1,IL18R1,SYNJ2BP 2.31E-09 1855 82

apoptosis

TAS2R10,EGFR,VCAM1,TNFSF11,IGF2,MAPK3,TG,CD5,NOS2,CSF3,AQP3,HBEGF,HS

PB1,CASP1,ATP2A1,APP,RETN,ILK,LPXN,FOLH1,E2F1,HSPD1,EGR1,TACR1,GJA5,FB

XO32,DUSP6,CXCL10,PTPRC,CXCL11,CXCR2,SELL,PPARD,WNT5A,IRF1,ALOX15,CX

CL1,PYCARD,SCARB1,DNM1,IRS2,IL1RN,IL7R,MMP7,FOXA2,TLR9,STAT1,TSC2,TXNIP,BCL2A1,ACVR2B,GNAI1,TP53INP1,FPR2,CPE,FAAH,GPX3,ATF4,VEGFC,MMP12,FC

GR3A,CFH,ICOS,BTG2,KLRK1,CTLA4,MAOA,IDO1,ORM1,VIPR1,KCNK5,CD86,CXCL9,

C4BPA,RCAN1,LMNA,OGDH,CA3,THRB,GPR44,SLURP1,SPRY2,GSTT1,TNFRSF13B,C

XCL13,BCL2L14,SCIN,NUPR1,MKL1,PCSK9,SHBG,CHI3L1,CXCL3,IBSP,PSMB9,PIGA,

RHOH,GZMM,CR2,EPHX1,EPM2A,GIMAP4,TAP1,SMAD6,CSRP1,NOV,MSRA,ENTPD5,

GRIA2,RASSF3,RIN2,CD74,BCL2L10,LBR,TSC22D1,FSCN1,TES,PARVA,TACC1,NEDD

9,PRPF31,GPR87,CSTA,MNDA,PLK4,GBP1,SFRP5,SMG1,GALR2,AKR7A2,UBD,PIK3IP

1,PACS2,PARP15,S100A6,TIMM8A,FILIP1L,CD3D,APOL1,AIM2,PINK1,MZF1,DEDD2,FA

IM2,SMPD3,HAGH,PAFAH2,PPM1A,SALL1,MOAP1,GRIN3A,PTPN7,OIP5,MYCT1,ACER

2,DMRT2,FREM1,SERPINB3,SERP2,CTRL,KCTD11,MUC17,DNASE1L1,TMEM109,CHA

C1 6.71E-08 5105 165

pregnancy

EGFR,VCAM1,TNFSF11,IGF2,MAPK3,TG,CD5,NOS2,CSF3,AQP3,HBEGF,HSPB1,CASP

1,RETN,HSPD1,EGR1,TACR1,GJA5,CXCL10,PTPRC,PPARD,IRF1,SCARB1,IL1RN,TLR

9,STAT1,TSC2,FAAH,ATF4,VEGFC,FCGR3A,BTG2,CTLA4,IL11RA,MAOA,IDO1,ORM1,T

HOP1,SHBG,PSMB9,EPHX1,SERPIND1,ERVWE1,HSD3B1,PECR,GALR2,S100A6,GC,D

ST,ST3GAL6,KLRC3,STOX1 1.12E-07 1052 52

T-cell response

VCAM1,TNFSF11,TG,CD5,NOS2,CSF3,CASP1,APP,FOLH1,E2F1,HSPD1,EGR1,CXCL10,PTPRC,SELL,PYCARD,IL7R,TLR9,STAT1,TXNIP,VEGFC,FCGR3A,ICOS,KLRK1,CTLA

4,IDO1,VIPR1,CD86,CXCL9,RCAN1,TNFRSF13B,TAP1,ENTPD5,CD74,BTN1A1,CLEC1

A,MAGEC2 3.30E-07 650 37

T-cell functionTNFSF11,MAPK3,CD5,NOS2,CSF3,E2F1,HSPD1,EGR1,CXCL10,PTPRC,SELL,PYCARD

,IL1RN,IL7R,TLR9,STAT1,AHNAK,ICOS,KLRK1,CTLA4,IDO1,VIPR1,KCNK5,CD86,TNFR

SF13B,LCP2,CR2,NEDD9,CD3D 8.81E-07 460 29

antigen processing and

presentationTNFSF11,TG,CD5,NOS2,HSPD1,CXCL10,PTPRC,IRF1,DNM1,IL7R,STAT1,FCGR3A,ICO

S,CTLA4,IDO1,CD86,THOP1,TNFRSF13B,PSMB9,CR2,TAP1,CD74,FSCN1,UBD,HLA-

DMA,HLA-DMB 1.07E-06 388 26

calcium mobilization

EGFR,IGF2,TACR3,CD5,HBEGF,APP,TACR1,CXCL10,PTPRC,CXCL11,CXCR2,SELL,W

NT5A,CXCL1,SCARB1,DNM1,TACR2,PRSS1,AHNAK,GNAI1,FPR2,FCGR3A,ICOS,KLR

K1,CTLA4,ORM1,CXCL9,GPR44,SPRY2,TNFRSF13B,CXCL13,SCIN,CXCL3,LCP2,CR2,

MCHR1,GALR2,RGS7,CD300E 1.36E-06 747 39

leukocyte migration

EGFR,VCAM1,MAPK3,CD5,NOS2,CSF3,HBEGF,HSPB1,APP,ILK,EGR1,CXCL10,CXCL1

1,CXCR2,SELL,PPARD,CXCL1,STAT1,FPR2,MMP12,ICOS,CTLA4,CD86,CXCL9,CXCL1

3,CXCL3,RHOH,VNN2 2.94E-06 462 28


17/33

Proteins regulating cell processes of inflammatory response


18/33

Copy Number Analysis


19/33

DNA Copy Number Analysis: Affy SNP 6.0 arrays

Raw Data Probe Level Copy Number: 1.6 million probes

Probe Level Segment Level Copy Number: 100K segments

Segment Level CIN Index: Whole Chromosome: 22Individual Cytobands: ~ 800


20/33

CIN Index - Cytoband level:

Relapse vs. Relapse_Free, t-test, P


21/33

Subnetwork enrichment for genes located in regions

with significant difference in copy number gains


22/33

Implication of copy number in CRC - Reference

Brosens et al. Cell Oncol (Dordr). 2011 Jun;34(3):215-23..Deletion of chromosome 4q predicts outcome in stage II colon cancer

patients.

RESULTS: Stage II colon cancers of patients who had relapse of

disease showed significantly more losses on chromosomes 4, 5, 15q,17q and 18q.

In the microsatellite stable (MSS) subgroup (n = 28), only

loss of cytobands 4q22.1- 4q35.2 was significantly associated with

disease relapse


23/33

Whole Exome Sequencing Variant Analysis

Pathway Level aggregation of variants: many related to


24/33

Pathway Level aggregation of variants: many related to

inflammation signaling


25/33

Data driven approach

Machine Learning Approaches:

- Support Vector Machines with Random

Feature Elimination (SVM-RFE)

- Random Forest for Integration and

prioritization (RF-ACE)

- Bayesian Networks


26/33

Classification Algorithm: Support Vector Machine (SVM)

Gene Expression

Tumors

microRNA

Expression

Tumors


27/33

Results of ROC Analysis for SVM classification

Data Type/Tissue Type Classifier AUC

Metabolites/Serum_Pos SVM 0.7

Gene

Expression/Tumors

SVM 0.81

miRNAExpression/Tumors SVM 0.84

CIN Index/Tumors SVM 0.87

Metabolites/Urine_Pos SVM 0.9


28/33

Regulome Explorer with Top Predictor-Target pairs

RF-ACE results: Top Features Associated With CRC Relapse


29/33

RF-ACE results: Top Features Associated With CRC Relapse

network viewer combination of genes, microRNA, copy number and

metabolites

Enrichment Analysis for RF-ACE top ranked features:


30/33

Enrichment Analysis for RF-ACE top ranked features:

Genes and microRNAs combined

They do converge!


31/33

Next Steps:

To refine systems biology findings Focusing on Inflammatory response pathways

Trends for co-expression pattern of specific

signaling relevant to regulatory T-lymphocytes

to narrow down validation experiments

To refine multivariate analysis using

predictive networks and downstream

systems biology What would be the best candidate biomarkers?


32/33

Acknowledgements:

Lombardi Comprehensive Cancer Center:

Genomics, Histopathology, and Metabolomics Shared

Resources

Analytical Group:

Krithika Bhuvaneshwar, Robinder Gauba, Lei Song, YuriyGusev

Design & Interpretation

Dr. Weiner, Dr. Byers, Dr. Gusev, Dr. Juhl

Innovation Center for Biomedical Informatics


33/33

Innovation Center for Biomedical Informatics

(ICBI)

Website : https://icbi.georgetown.edu

Email : [email protected]
https://informatics.georgetown.edu/mailto:[email protected]:[email protected]://informatics.georgetown.edu/

platform for personalized oncology

Documents