Pistoia Alliance US Conference 2015 - 1.5.4 New data - Nikolaus Schultz

Nikolaus Schultz

Marie Josée and Henry R. Kravis Center for Molecular Oncology

Memorial Sloan Kettering Cancer Center

October 13, 2015

Visualization and Analysis of Cancer Genomics Data

Cost of DNA Sequencing is dropping rapidly

The Hallmarks of Cancer

Hanahan and Weinberg. Cell. March 4 2011.

Cancer is a class of diseases in which a group of cells display: uncontrolled growth invasion that intrudes upon and destroys adjacent tissues, and sometimes metastasis (spreading to other locations in the body via lymph or blood)

Many of these mechanisms are known, many not. Only some are treatable.

All these properties are caused by genetic or epigenetic alterations.

Can we identify the responsible alterations in the genomes of cancer patients?

The Hallmarks of CancerCancer is a class of diseases in which a group of cells display: uncontrolled growth invasion that intrudes upon and destroys adjacent tissues, and sometimes metastasis (spreading to other locations in the body via lymph or blood)

Tumor development / Drivers versus passengers

How does a cancer cell acquire all these different alterations?

Sequential accumulation of genomic alterations that confer a growth advantage (like in evolution, but faster).

Certain early events can increase the rate of accumulation, like mutations in DNA damage repair genes or cell-cycle checkpoint genes (or mutagens).

Over time, many alterations develop. The ones that confer a growth advantage are called “drivers”, all others are “passengers”. Can we distinguish between them?

Identification of functional alterations in genomic data - per disease - per gene - per patient - per pathway

Different, recurrent ways to alter the same pathway / process?

Many events are rare, so we need hundreds of samples of the same disease (sub-)type to find them based on recurrence!

Clinical applications:Development of new prognostic toolsIdentification of new treatment options

Patient-specific treatment

Utility of cancer genomics data

bioinformaticians

biologists

clinicians

2010 2011 2012 2013

Kidney clear cell

Endometrial cancer

Thyroid cancer

Head & neck squamous

Lung squamous cell carcinoma

Colorectal cancer

Breast cancer

Low grade glioma

2014

GBM Phase II

Bladder cancer

Lung adenocarcinoma

Melanoma

Prostate cancer

Stomach adenocarcinoma

+ lobular breast cancer, chromophobe kidney, papillary kidney,, pancreatic, rare tumors …500 samples per tumor type 10,000 tumor / normal pairs total

The Cancer Genome Atlas Project History20092008

Ovarian cancerGBM

AML

Cervical

Liver

Sarcoma

2015

Cancer Cell Line Encyclopedia (CCLE)

Broad Institute, Sanger, Washington University, etc.

Tumor sequencing in hospitals (MSKCC 500 per month)

Sources of tumor sequencing data

10,000 tumors

6,000 tumors

1,000 cell lines

5,000 tumors

>15,000 tumors

Raw data (FASTQ / BAM files)dbGaP, CGHub, ICGC Data Portal

Processed data (gene level data, mutation calls)

TCGA Data Portal, ICGC Data Portal, Supplementary Tables

Data slices (subsets of processed data)

Data visualization and analysis tools

Data availability

bioinformaticians

biologists, clinicians

Raw data (FASTQ / BAM files)dbGaP, CGHub, ICGC Data Portal

Processed data (gene level data, mutation calls)

TCGA Data Portal, ICGC Data Portal, Supplementary Tables

Data slices (subsets of processed data)

Data visualization and analysis tools

Data availability

bioinformaticians

biologists, cliniciansReduction of complexity!

Most mutations found in cancer are “passengers”

Driver alteration frequencies per tumor type

Driver alteration frequencies per tumor type

Rec L domain Furin-like Rec L domain Kinase domain

ERBB2 mutation hotspots across cancer types

Rec L domain Furin-like Rec L domain Kinase domain

ERBB2 mutation hotspots across cancer types

signal

noise

ERBB2 mutation hotspots across cancer typesS310F

Bladder: 1Breast: 3Cervical: 1Colorectal: 2Lung adeno: 2Ovarian: 2 Stomach: 1CCLE: 1 (bladder)

L755S/M/P/WBreast: 4Colorectal: 2Endometrial: 1Kidney (pap): 1Melanoma: 1Stomach: 1CCLE: 3 (colorectal, stomach, brain)

V777L/ABreast: 1Colorectal: 2GBM: 2

V842IBreast: 1Colorectal: 4Endometrial: 2CCLE: 4 (Lung, ovarian, endometrial)

R678QBreast: 1Colorectal: 1Endometrial: 1Stomach: 2CCLE: 1 (colorectal)

774-776insLung adeno: 6CCLE: 1 (lung)

Rec L domain Furin-like Rec L domain Kinase domain

ERBB2 mutation hotspots across cancer typesS310F

Bladder: 1Breast: 3Cervical: 1Colorectal: 2Lung adeno: 2Ovarian: 2 Stomach: 1CCLE: 1 (bladder)

L755S/M/P/WBreast: 4Colorectal: 2Endometrial: 1Kidney (pap): 1Melanoma: 1Stomach: 1CCLE: 3 (colorectal, stomach, brain)

V777L/ABreast: 1Colorectal: 2GBM: 2

V842IBreast: 1Colorectal: 4Endometrial: 2CCLE: 4 (Lung, ovarian, endometrial)

R678QBreast: 1Colorectal: 1Endometrial: 1Stomach: 2CCLE: 1 (colorectal)

774-776insLung adeno: 6CCLE: 1 (lung)

Rec L domain Furin-like Rec L domain Kinase domain

Greulich et al. PNAS 2012.

Kancha et al. PLoS ONE 2011.

Bose et al. Cancer Discovery 2012.

Bose et al. Cancer Discovery 2012.

Bose et al. Cancer Discovery 2012.

cBioPortal for Cancer Genomics: Data to knowledge

Tumor DNA DNA sequencer, microarrays …

Tumor and normalsequences

Data

Intuitive interface, quick response time, reduction of complexity

Alteration types and thresholds can be customized for each gene.

Reduction of complexity: Event callsWhich genes are altered in which samples?

cBioPortal

Data visualization and exploration in cBioPortal

ClinicalMSK-IMPACT

Geno

mic

dat

a

CMO Research

FoundationMedicine

ClinicalMSK-IMPACT

Geno

mic

dat

a

CMO Research cBioPortal

Data visualization and exploration in cBioPortal

TCGA, ICGC

Other public data

FoundationMedicine

ClinicalMSK-IMPACT

Geno

mic

dat

a

CMO Research cBioPortal

FoundationMedicine

TCGA, ICGC

Other public data

MSKCCclinical data

Data visualization and exploration in cBioPortal

ClinicalMSK-IMPACT

Geno

mic

dat

a

CMO Research cBioPortal

OncoKB: Annotation of variant effects, treatment

FoundationMedicine

TCGA, ICGC

Other public data

Clinical annotationStep 1: ManualStep 2: Automated via

institutional databases

MSKCCclinical data

OncoKBKnowledgebase

of oncogenic mutations

Variant effectNCCN guidelines

Standard therapy

Investigationaltherapy

Clinical trials

Live demonstration of cBioPortal

http://cbioportal.org/

cBioPortal usage and interest cbioportal.org

>5,000 unique users per week, doubling every year

cBioPortal usage and interest cbioportal.org

>5,000 unique users per week, doubling every year

Numerous academic installations of cBioPortal:Dana-Farber, Princess Margaret, CHOP, Weill Cornell, Fred Hutchinson, UCSC, Columbia, NYU, NY Genome Center, British Columbia, University of Michigan, SickKids, Vanderbilt, Emory, UNC, University of Pittsburgh, CRUK, EMBL, Charite Berlin, institutions in Japan, China, …

Interest by several people to modify or customize the code, and to contribute new features

Interest by pharmaceutical companies and others to use cBioPortal● For internal data analysis (large pharma)● In customer-facing applications (smaller service companies)

Switch to open sourcecBioPortal source code is available via GitHub:

https://github.com/cBioPortal/cbioportalAGPL license v3 (Affero GPL):

A GPL variant, main difference is that redistribution over a network triggers the copyleft requirements

Impact on cancer research, patient treatment, drug development through:• More robust and flexible software• Accelerated development of new features• Wider user base, collaborative culture

Core cBioPortal Development groupMemorial Sloan Kettering Cancer Center

Nikolaus Schultz, Chris Sander, Benjamin Gross, JJ Gao

Dana Farber Cancer InstituteEthan Cerami

Princess Margaret Cancer Centre Trevor Pugh, Stuart Watt

Re-uniting two cBioPortal foundersCoordination of architectural decisions, feature development, merges, etc.

TheHyve now offering commercial services around cBioPortal

Summary

Rapidly growing body of cancer genomics data (public and private)

Reduction of complexity can make these data accessible and interpretable

cBioPortal allows access to cancer genomics data sets:cbioportal.org: public sitevia GitHub: install local versions

cBioPortal is now fully open sourcesoftwaredata pipelines and data sets coming sooncommercial support available

Still exploring pre-competitive funding options

Acknowledgements

CMOCyriac KandothWilliam LeeRajmohan MuraliNicholas D. SocciBarry TaylorMichael BergerAgnes VialeDavid B. SolitMichael Trapani Ederlinda Paraiso

Molecular DiagnosticsAhmet ZehirAijaz SyedDonavan ChengMichael BergerMaria ArcilaMarc LadanyiInformation SystemsMike EubanksStu Gardos

cBioPortalJianJiong GaoBenjamin GrossYichao SunHongxin ZhangFred CriscuoloDong LiAdam AbeshouseRitika KundraAnnice Chen

Chris SanderOnur SumerArman AksoyEthan Cerami

KnowledgebaseDebyani ChakravartySarah PhillipsJulia Rudolph

Bioinformatics CoreJoanne Edington

Demo slides

End of Live Demo