CELL INDEX DATABASE (CELLX): A WEB TOOL FOR CANCER PRECISION MEDICINE *

KEITH A. CHING1, KAI WANG1, ZHENGYAN KAN1, JULIO FERNANDEZ1, WENYAN ZHONG1, JAREK KOSTROWICKI1, TAO XIE1, ZHOU ZHU1, JEAN-FRANCOIS MARTINI2, MARIA KOEHLER2, KIM ARNDT1,

PAUL REJTO1

1Oncology Research Unit, 2Oncology Business Unit, Pfizer Global Research & Development, Pfizer Inc., 10777 Science Center Drive San Diego, CA 92121, USA Email: keith.ching@pfizer.com

The Cell Index Database, (CELLX) (http://cellx.sourceforge.net) provides a computational framework for integrating expression, copy number variation, mutation, compound activity, and meta data from cancer cells. CELLX provides the computational biologist a quick way to perform routine analyses as well as the means to rapidly integrate data for offline analysis. Data is accessible through a web interface which utilizes R to generate plots and perform clustering, correlations, and statistical tests for associations within and between data types for ~20,000 samples from TCGA, CCLE, Sanger, GSK, GEO, GTEx, and other public sources. We show how CELLX supports precision oncology through indications discovery, biomarker evaluation, and cell line screening analysis.

1. Introduction

To support precision medicine patient selection strategies, genomics data is used to identify oncogenic drivers or dysregulated pathways in cancer cells susceptible to therapeutic intervention. Notably, efforts by The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov), the Cancer Cell Line Encyclopedia (CCLE)[1], and Sanger Wellcome Trust Genomics of Drug Sensitivity in Cancer (GDSC)[2] have generated a plethora of data and datatypes that can be used for generating patient selection hypotheses. However, multiple genomics data types such as expression, copy number variation (CNV), and mutation are large and unwieldy to manage. For the computational biologist, much time and effort can be spent to assemble an up to date table of features which can be computed on because new data are often generated frequently and incrementally. Thus, there is a need for an infrastructure to perform simple, quick, and routine analyses on multi-dimensional genomics data as well as the automated assembly of data tables for offline computation using more sophisticated algorithms. !Currently, there exist several cancer genomics databases to access expression, CNV, mutation, and integrated data as reviewed in [3]. For example, BioGPS[4] provides expression data, Tumorscape[5] contains CNV measurements, the Sanger Catalog of Somatic Mutations in Cancer (COSMIC)[6] lists mutations, and the cBio Portal[7] integrates multiple TCGA data types. Additionally, databases with compound activity data include GDSC and CCLE. Here we present a publicly available web-based informatics tool to integrate data, perform analysis, and visualize results from public as well as private internal sources to support precision medicine activities. !

* This work is supported by Pfizer, Inc.

2. Architecture

The underlying MySQL database consists of 22 tables for expression, CNV, mutation, compound, sample, meta data, RNAi, RPPA, and gene annotation data. The Perl CELLX application runs on an Apache web server. R-serve (http://www.rforge.net/Rserve/) instances generate plots and perform statistical analyses. An Apache Tomcat application server runs a custom Java servlet which bridges Perl and R by funneling Perl http requests to the R-serves and sends results back to the web server. A demo site, instructions, source code, database dumps, and data parsing / loading scripts are available at http://cellx.sourceforge.net.

3. Gene Based Search

A common starting point for indications discovery is asking where the target of interest is altered. CELLX can plot the relative expression or CNV of a gene within a dataset or across multiple compatible datasets. For instance, RNA-Seq data processed by RSEM[8] can be compared across tumors profiled not only by TCGA, but CCLE as well. CDK4 expression can be seen to have high outliers in Glioblastoma Multiforme (GBM), melanoma (SKCM), breast (BRCA), Lower Grade Glioma (LGG), and sarcomas (SARC) (Figure 1). A similar plot can be generated of CNV to identify datasets with amplifications or deletions. CELLX can chart the relationship between expression and CNV across datasets using scatter plots of expression versus CNV. A hallmark of amplification, CDK4 expression levels scale with CNV level in several datasets (Figure 2a,b). !!!!!!!!!!!!!!!!!!!!!!!!!!!!

CCLE−R

SEM

TCGA−AC

C−RS

EMTC

GA−BLCA

−RSE

MTC

GA−BLCA

−RSE

M_N

TCGA−BR

CA−R

SEM

TCGA−BR

CA−R

SEM_N

TCGA−CE

SC−R

SEM

TCGA−CE

SC−R

SEM_N

TCGA−CO

AD−R

SEM

TCGA−CO

AD−R

SEM_N

TCGA−DL

BC−R

SEM

TCGA−GBM

−RSE

MTC

GA−HN

SC−R

SEM

TCGA−HN

SC−R

SEM_N

TCGA−KICH

−RSE

MTC

GA−KICH

−RSE

M_N

TCGA−KIRC

−RSE

MTC

GA−KIRC

−RSE

M_N

TCGA−KIRP

−RSE

MTC

GA−KIRP

−RSE

M_N

TCGA−LAML−RS

EMTC

GA−LG

G−R

SEM

TCGA−LIHC

−RSE

MTC

GA−LIHC

−RSE

M_N

TCGA−LUAD

−RSE

MTC

GA−LUAD

−RSE

M_N

TCGA−LU

SC−R

SEM

TCGA−LU

SC−R

SEM_N

TCGA−MES

O−R

SEM

TCGA−OV

−RSE

MTC

GA−PAAD

−RSE

MTC

GA−PAAD

−RSE

M_N

TCGA−PC

PG−R

SEM

TCGA−PC

PG−R

SEM_N

TCGA−PR

AD−R

SEM

TCGA−PR

AD−R

SEM_N

TCGA−RE

AD−R

SEM

TCGA−RE

AD−R

SEM_N

TCGA−SA

RC−R

SEM

TCGA−SA

RC−R

SEM_N

TCGA−SK

CM−R

SEM

TCGA−SK

CM−R

SEM_N

TCGA−TH

CA−R

SEM

TCGA−TH

CA−R

SEM_N

TCGA−UC

EC−R

SEM

TCGA−UC

EC−R

SEM_N

TCGA−UC

S−RS

EM

0

5

10

15

RSEM Expression CDK4

CDK4

TumorNormal

Figure 1. RNA-Seq RSEM gene expression of CDK4 (y-axis, log2) across datasets shows higher expression in tumor vs. adjacent normal tissue. Particular groups of outliers can be seen in GBM (glioblastoma multiforme), SARC (sarcoma), SKCM (skin cutaneous melanoma), LGG (brain lower grade glioma), and cell lines (CCLE).

!!!!!!!!!!!!

4. Integrated Visualization

Mixed data types can be visualized in 2D scatter plots to look at the relationship between two datatypes on the same or different genes. For instance, expression of gene A on the x-axis can be plotted versus the CNV of gene B on the y-axis. Other plottable datatypes are protein levels for Reverse Phase Protein Arrays (RPPA), the mutation count per sample, the general amount of CNV per sample, IC50 values for compounds, and meta data. Multiple layers of data can be added to the plot to increase dimensionality. As a simple example, one can plot the expression of ERBB2 expression vs. ERBB2 CNV overlaid with ERBB2 mutations (Figure 3a) or breast cancer subtype meta data. (Figure 3b). The underlying data used to generate each plot is linked as a tab separated tsv file for downloading. !!!!!!!!!!!

5 10 15

810

1214

1618

Expression TCGA−BRCA−RSEM

ESR1

ERBB

2

BasalBasal

BasalBasal

Basal

Basal Basal

Basal

Basal

BasalBasal

BasalBasal

Basal

Basal

Basal

BasalBasalBasal

Basal

Basal

Basal

Basal

Basal

Basal

Basal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

BasalBasal

Basal

Basal

BasalBasal

Basal

BasalBasal

Basal

BasalBasal

BasalBasal

Basal

Basal

Basal

Basal

BasalBasal

Basal

BasalBasal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal

Basal

Basal

BasalBasal

Basal

BasalBasal

Basal

Basal

BasalBasalBasal

Basal

Basal

BasalBasal

BasalBasal

BasalBasal

BasalBasal

Basal

Basal

Basal

Basal

Basal Basal

Basal

Basal

BasalBasal Basal

BasalBasal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

BasalBasal

BasalBasal

Basal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal

Basal

Basal

Basal BasalBasal

Basal

Basal

Basal BasalBasal

Basal

Basal

Basal

Basal

BasalBasal

Basal

Basal

Basal

Basal Basal

Basal

Basal

BasalBasalBasal

Basal

Her2

Her2 Her2

Her2Her2

Her2

Her2

Her2Her2 Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2 Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2

Her2Her2 Her2

Her2

Her2

Her2

Her2Her2Her2

Her2

Her2

Her2Her2

Her2Her2

Her2Her2

Her2

Her2

Her2

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumALumA

LumA

LumA

LumALumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumALumALumALumA LumA

LumALumA

LumALumA LumA

LumA

LumALumA

LumA

LumALumALumA

LumALumALumA

LumA

LumA

LumA

LumALumA

LumALumA

LumA

LumALumALumA

LumA

LumA

LumALumA

LumA

LumA LumA

LumA

LumALumA LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumALumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA LumALumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA LumALumA

LumA

LumA

LumA

LumA

LumA

LumALumALumALumA

LumA

LumA

LumA

LumALumA

LumA

LumALumALumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumALumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumALumALumA

LumALumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumALumA

LumALumA

LumA LumA

LumA

LumALumA

LumALumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA LumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumALumALumA

LumA

LumALumA

LumALumA

LumA

LumALumA LumA

LumA LumA

LumA

LumA

LumALumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumALumA

LumA

LumALumA

LumA

LumALumA

LumA

LumA

LumALumA LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumALumA

LumA

LumALumALumALumA

LumA

LumA

LumA

LumALumA

LumA

LumALumA LumA

LumALumA

LumALumALumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumA

LumALumALumA

LumA

LumA

LumA

LumALumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA LumA

LumA

LumALumA

LumA

LumALumA

LumA

LumALumA

LumALumA

LumALumALumA

LumA

LumA

LumA

LumALumALumA LumA

LumA

LumA

LumA

LumALumALumA

LumALumA

LumA

LumA

LumA

LumA

LumA

LumA

LumA

LumALumALumA

LumA

LumA

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumBLumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumBLumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumBLumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumBLumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumBLumB

LumB LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB LumB LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumBLumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB LumB

LumB

LumB

LumBLumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumBLumB

LumB

LumB

LumB

LumB

LumBLumBLumB

LumBLumB

LumB

LumB

LumB

LumBLumBLumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB LumB

LumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumBLumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumB

LumBLumB

LumB

LumB

LumB

LumB

LumB

meta_PAM50.RNASEQ

5 10 15

810

1214

1618

Expression TCGA−BRCA−RSEM

ESR1

ERBB

2

foc foc

foc

focfoc

foc

focfoc

foc

foc

foc foc

foc

foc

foc

foc

foc

foc

focfoc

foc

foc

foc

foc

foc

foc

foc

foc

focfoc

foc

foc

foc foc

foc

foc

focfoc

focfoc

foc

foc

foc

foc

foc

focfoc

foc

foc

focfocfoc

focfoc

foc foc

foc

foc

foc

focfoc foc

foc

foc

foc

focfoc

focfoc

foc

foc

foc focfoc

foc

foc

focfocfoc

foc

focfocfoc

focfoc

foc

foc

focfoc

foc

mutantAmp >= 1Del <= −2FocalERBB2_cERBB2_m

Figure 2. Correlation of expression and CNV. CNV (y-axis in log2 diploid genome) vs. RSEM expression levels (log2) for CDK4 show that a) SARC and b) GBM datasets have a sizable population of cells overexpressing CDK4 due to amplification of the locus. Additionally, expression levels scale with CNV levels. Clear outliers from the main distribution of CNV values can help determine appropriate CNV cut offs for amplification status. In this example, samples colored red have > 1 log2 diploid genomes (i.e. >~4 copies).

Figure 3. 2D scatter plots. a) Gene expression of ESR1 (x-axis, log2) vs. ERBB2 (y-axis, log2) gene expression. ERBB2 CNV over the selected threshold of 1 (log2 diploid genome) is colored pink. Focal amplifications (< 10MB) are denoted with ‘foc’. Mutations in ERBB2 are colored green. c) Meta data for PAM50 subtype classification are colored and overlaid on the ESR1 vs. ERBB2 gene expression plot.

a) b)

11 12 13 14 15 16

−2−1

01

2

TCGA−LUSC−RSEM

CDK4

CDK4_c

11.0 11.5 12.0 12.5 13.0

−2−1

01

2

TCGA−MESO−RSEM

CDK4

CDK4_c

10 11 12 13

−2−1

01

2

TCGA−OV−RSEM

CDK4

CDK4_c

10.5 11.0 11.5 12.0

−2−1

01

2

TCGA−PAAD−RSEM

CDK4

CDK4_c

10.5 11.0 11.5 12.0 12.5

−2−1

01

2

TCGA−PCPG−RSEM

CDK4

CDK4_c

9.5 10.0 10.5 11.0 11.5 12.0

−2−1

01

2

TCGA−PRAD−RSEM

CDK4

CDK4_c

10.5 11.0 11.5 12.0 12.5 13.0

−2−1

01

2

TCGA−READ−RSEM

CDK4

CDK4_c

10 12 14 16

−2−1

01

2

TCGA−SARC−RSEM

CDK4

CDK4_c

10 11 12 13 14 15 16 17

−2−1

01

2

CCLE−RSEM

CDK4

CDK4_c

10.5 11.0 11.5 12.0 12.5 13.0 13.5

−2−1

01

2

TCGA−ACC−RSEM

CDK4

CDK4_c

9 10 11 12 13

−2−1

01

2

TCGA−BLCA−RSEM

CDK4

CDK4_c

9 10 11 12 13 14 15

−2−1

01

2

TCGA−BRCA−RSEM

CDK4

CDK4_c

10.5 11.0 11.5 12.0 12.5 13.0

−2−1

01

2

TCGA−CESC−RSEM

CDK4

CDK4_c

9 10 11 12 13 14

−2−1

01

2

TCGA−COAD−RSEM

CDK4

CDK4_c

11.0 11.5 12.0 12.5 13.0

−2−1

01

2

TCGA−DLBC−RSEM

CDK4

CDK4_c

12 14 16 18

−2−1

01

2

TCGA−GBM−RSEM

CDK4

CDK4_c

a) b)

5. Biomarker Frequency Reports

Tables of the frequency of alterations across datasets can help to prioritize indications for therapies with known biomarkers. For instance, the venn report of the frequency of CDK4 biomarker alterations within datasets shows significant frequencies of CDK4 amplification in sarcoma, gliomas, and melanoma TCGA datasets (Table 1). Cutoffs can be defined by expression level, CNV level, and/or mutation status. The co-occurrence or exclusion of 2-4 biomarkers within the same sample can also be quantified.

6. Analysis

CELLX can identify genes whose expression correlates with a gene of interest and return a table of significant genes that can be visualized via a heat map with labelled metadata. For example, a search for genes correlated with CDK4 expression in the TCGA sarcoma dataset yields ACVRL1 which is expressed by vascular endothelium and a potential anti-angiogenesis target. (Figure 4a)

sourcename CDK4_c cells_c CDK4_m cells_m cell_type tumor_type CNV% MUT%TCGA-SARC 35 171 0 0 soft_tissue Sarcoma 20.47 NATCGA-GBM 73 607 0 150 neuronal Glioblastoma multiforme 12.03 0TCGA-LGG 14 471 1 612 neuronal Brain Lower Grade Glioma 2.97 0.16TCGA-ACC 2 90 0 91 adrenal_gland Adrenocortical carcinoma 2.22 0TCGA-SKCM 7 387 8 372 skin Skin Cutaneous Melanoma 1.81 2.15TCGA-LUAD 5 510 3 491 lung Lung adenocarcinoma 0.98 0.61TCGA-STAD 2 403 1 373 stomach Stomach adenocarcinoma 0.5 0.27TCGA-BRCA 5 1074 1 777 breast Breast invasive carcinoma 0.47 0.13TCGA-BLCA 1 255 2 242 urinary_tract Bladder Urothelial Carcinoma 0.39 0.83TCGA-OV 2 569 0 476 ovary Ovarian serous cystadenocarcinoma 0.35 0TCGA-LUSC 1 487 0 233 lung Lung squamous cell carcinoma 0.21 0TCGA-COAD 0 446 2 219 large_intestine Colon adenocarcinoma 0 0.91TCGA-PRAD 0 381 0 300 prostate Prostate adenocarcinoma 0 0TCGA-THCA 0 508 0 428 thyroid Thyroid carcinoma 0 0TCGA-PAAD 0 92 1 91 pancreas Pancreatic adenocarcinoma 0 1.1TCGA-PCPG 0 175 0 0 adrenal_gland Pheochromocytoma and Paraganglioma 0 NATCGA-MESO 0 37 0 0 pleura Mesothelioma 0 NATCGA-READ 0 164 0 1 rectum Rectum adenocarcinoma 0 0TCGA-UCEC 0 533 5 248 endometrium Uterine Corpus Endometrial Carcinoma 0 2.02TCGA-KIRC 0 521 6 328 kidney Kidney renal clear cell carcinoma 0 1.83TCGA-ESCA 0 126 0 0 oesophagus Esophageal carcinoma 0 NATCGA-DLBC 0 28 0 79 haematopoietic_

and_lymphoid_tiLymphoid Neoplasm Diffuse Large B-cell Lymphoma

0 0TCGA-KICH 0 66 0 66 kidney Kidney Chromophobe 0 0TCGA-UCS 0 57 0 57 uterus Uterine Carcinosarcoma 0 0TCGA-KIRP 0 212 0 169 kidney Kidney renal papillary cell carcinoma 0 0TCGA-LAML 0 194 0 118 haematopoietic_

and_lymphoid_tiAcute Myeloid Leukemia 0 0

TCGA-LIHC 0 213 5 202 liver Liver hepatocellular carcinoma 0 2.48TCGA-HNSC 0 516 5 513 upper_aerodiges

tive_tractHead and Neck squamous cell carcinoma 0 0.97

TCGA-CESC 0 206 0 41 cervix Cervical squamous cell carcinoma and endocervical adenocarcinoma

0 0

Table 1. Frequency report for CDK4 alterations in TCGA. CDK4_c is the number of samples in which the CNV exceeds the set threshold, in this case ~4 copies. CDK4_m is the number of samples with a CDK4 mutation. The cells_c/_m columns are the number of samples for which CNV or mutation data are available, respectively. Percentages are calculated as altered / total for each individual alteration type.

!!!!!!!!!!! !

!!!!!!!

A scatter plot of CDK4 vs. ACVRL1 shows higher ACVRL1 in Dedifferentiated Liposarcomas (DDPLS) vs. Leiomyosarcomas (Figure 4b). This is consistent with a study reporting immature and intermediate blood vessels in sarcomas and quantifying tumor microvessel density that is ~3X higher in DDLPS vs. Leiomyosarcomas. [9] The plot also shows that CDK4 expression is high in DDLPS and often focally amplified which is consistent with the literature.[10] CELLX can also

metavalue min_pvalhistologic_diagnosis 2.54E-19well_differentiated_liposarcoma_primary_dx 5.62E-08residual_tumor 2.88E-05leiomyosarcoma_uterine_involvement 0.010166952gender 0.011140224histologic_subtype 0.012659498primary_tumor_lower_uterus_segment 0.030764968prior_dx 0.033818919history_of_neoadjuvant_treatment 0.043692764

10 11 12 13 14 15 16 17

68

1012

14

Expression TCGA−SARC−RSEM

CDK4

ACVR

L1

foc

foc

foc

foc

foc

foc

foc

foc

foc

foc

foc

foc

foc

focfoc

foc

foc

foc

foc

foc

foc

foc

foc

foc

focfoc

foc

foc

Ded

Ded

Ded

Ded

Ded

Ded

Ded

DedDed

Ded

Ded

Ded

Ded

Ded

DedDed

DedDed

Ded

Ded

Ded

Ded

Ded

Ded

Ded

Ded

Ded

Ded

DedDed

Ded

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

LeiLei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei

Lei Lei

Lei

Lei

LeiLei

LeiLei

Lei

Lei

Lei

Lei Lei

Lei

Lei

Lei

Myx

Myx

Und

Und

Und

Und

Und

Amp >= 1Del <= −2FocalCDK4_cmeta_histologic_diagnosis

Figure 4. Analysis of features associated with CDK4 expression. a) Heatmap of top 200 genes (columns) correlated with CDK4 expression levels in samples (rows) from the TCGA sarcoma dataset showing ACVRL1 expression correlates with CDK4 (arrows). Meta data labels for histologic diagnosis are colored in a column on the left side of the plot. b) Scatter plot of CDK4 expression versus ACVRL1 expression showing high ACVRL1 expression in dedifferentiated liposarcomas. Metavalues from a) are colored and abbreviated by the first 3 letters. Amplification of CDK4 is denoted by a violet circle. foc=focal. c) Meta data with significantly different CDK4 expression levels. Min p-value is the lowest pairwise t-test score. d) Boxplot of histologic diagnosis by CDK4 expression data used in c).

SPAR

CCO

L1A1

COL3

A1M

MP2

FBN1

COL5

A2CO

L5A1

OS9

LRP1

FSTL

1DC

NM

XRA5

CD24

8H1

9CC

T2CD

K4CT

DSP2

RNAS

E1M

RC2

MAR

CKS

CHPF

CKAP

4PC

DHG

C3TS

PAN3

1M

DM2

MFA

P2PT

K7CE

RCAM FY

NDA

B2TT

YH3

FSCN

1M

AFCO

PZ1

TMEM

214

SNX1

7NR

BP1

HDAC

7SA

R1A

SEP1

5RA

B1A

TM9S

F3C4

BPB

KCNM

B2CY

P2W

1HY

MAI

LOC3

8903

3KC

NG2

ZG16

BHM

GA2

TNNT

3SG

CGAD

H1C

SMTN

L2KA

NK4

DLK1

MEG

3XP

NPEP

2TN

FAIP

6PL

AC9

HSD1

1B2

COLG

ALT2

IGDC

C4IL

17D

RFTN

2KB

TBD1

1G

AP43

SPAT

A18

GRI

N2D

CYP2

7B1

AVIL

CR1

FHDC

1TN

FRSF

8G

PR85

C19o

rf38

KRT2

22PD

E9A

RPH3

ALPD

PNPT

GES

MED

AGLY

VE1

CCDC

8AD

AMTS

14KC

NMB4

APO

L4CD

C42E

P5PR

R5AG

AP2

GBG

T1KC

NQ1

IGF1

CCND

2PL

AGL1

ADAM

TS2

HSPA

12B

S1PR

2PA

PSS2

TFPI

SPRY

1FG

D5PC

DH12

GAL

CAR

HGAP

18G

CNT1

XRCC

6BP1

SLC3

5E3

ADA

FKBP

11SL

C18B

1TB

PL1

ABRA

CLFK

BP14

PLEK

HG1

PRKC

EZD

HHC1

FAM

45A

PTRH

D1TO

R2A

YEAT

S4Q

SOX2

GXY

LT1

NRBF

2SL

C17A

5AC

VR1B

C10o

rf32

NDST

2DH

X32

PPP2

R2D

SUFU

MAR

CH5

PYCR

1CP

XM1

LAM

A2PD

GFR

ATM

EM11

9CR

EB3L

1CO

LEC1

2AG

AP2.

AS1

MAR

CH9

MET

TL21

BM

ETTL

1TS

FMG

NG2

JDP2

ASAP

3AD

D3G

PX7

SESN

1FK

BP7

C11o

rf95

B3G

NT9

UBE2

E3C9

orf6

9NC

KAP5

LDN

AJB1

2NT

5C2

XPNP

EP1

CREM

LTV1

REPS

1SE

SN2

MPD

U1EM

L4SL

C35C

1TM

EM9B

ZDHH

C9AD

PGK

SRPX

FRS2

DYRK

2RG

L1CN

OT2

LOC6

5434

2AC

VRL1

MET

RNL

LSP1

TNFR

SF1B

FUCA

2IF

NGR1

GO

LM1

CMTM

3PG

DEF

NB1

P4HA

1LE

PREL

4LI

MK1

FAM

3CG

PX8

WBP

1LRB

MS1

NECA

P2LI

PATS

PAN1

4AL

DH18

A1O

STC

IFNG

R2C5

orf1

5G

LT8D

1DC

TD

TCGA−HS−A5N9−01ATCGA−HB−A43Z−01ATCGA−HB−A3L4−01ATCGA−IS−A3KA−01ATCGA−IS−A3K6−01ATCGA−IE−A4EH−01ATCGA−KD−A5QU−01ATCGA−K1−A42X−02ATCGA−K1−A42X−01ATCGA−HS−A5NA−01ATCGA−KD−A5QT−01ATCGA−DX−A3UC−01ATCGA−DX−A48R−01ATCGA−IE−A4EI−01ATCGA−IE−A4EK−01ATCGA−DX−A6B7−01ATCGA−DX−A3UD−01ATCGA−DX−A48U−01ATCGA−DX−A48O−01ATCGA−MO−A47P−01ATCGA−MB−A5YA−01ATCGA−K1−A42W−01ATCGA−PC−A5DL−01ATCGA−DX−A48P−01ATCGA−HS−A5N7−01ATCGA−DX−A3UF−01ATCGA−FX−A3NJ−01ATCGA−PC−A5DK−01ATCGA−DX−A48J−01ATCGA−PC−A5DO−01ATCGA−DX−A3U7−01ATCGA−DX−A3U9−01ATCGA−IE−A3OV−01ATCGA−DX−A3UE−01ATCGA−DX−A48L−01ATCGA−IF−A4AJ−01ATCGA−MJ−A68H−01ATCGA−IF−A4AK−01ATCGA−K1−A3PN−01ATCGA−K1−A3PN−02ATCGA−DX−A6BA−01ATCGA−PC−A5DP−01ATCGA−HB−A5W3−01ATCGA−DX−A3LY−01BTCGA−DX−A3U5−01ATCGA−DX−A3LS−01ATCGA−DX−A1KU−01ATCGA−DX−A1KZ−01ATCGA−DX−A23Y−01ATCGA−DX−A3LT−01ATCGA−DX−A3U6−01ATCGA−DX−A23T−01ATCGA−DX−A2J1−01ATCGA−DX−A2J4−01ATCGA−DX−A1L3−01ATCGA−DX−A3LU−01ATCGA−IF−A3RQ−01ATCGA−DX−A23V−01ATCGA−DX−A23Z−01ATCGA−DX−A2J0−01ATCGA−DX−A1L0−01ATCGA−DX−A23R−01ATCGA−DX−A3M1−01ATCGA−DX−A3LW−01ATCGA−HB−A3YV−01ATCGA−DX−A48N−01ATCGA−DX−A1KW−01ATCGA−DX−A6BH−01ATCGA−FX−A3NK−01ATCGA−RN−A68Q−01ATCGA−DX−A2IZ−01ATCGA−MO−A47R−01ATCGA−FX−A2QS−01ATCGA−FX−A3RE−01ATCGA−PC−A5DN−01ATCGA−IW−A3M5−01ATCGA−HS−A5N8−01ATCGA−DX−A23U−01ATCGA−JV−A5VF−01ATCGA−IS−A3K7−01ATCGA−DX−A1L2−01ATCGA−DX−A1L1−01ATCGA−MJ−A68J−01ATCGA−MB−A5Y9−01ATCGA−DX−A3M2−01ATCGA−FX−A3TO−01ATCGA−IE−A6BZ−01ATCGA−IE−A4EJ−01ATCGA−DX−A1KX−01ATCGA−DX−A3UB−01ATCGA−K1−A3PO−01ATCGA−FX−A48G−01ATCGA−IW−A3M4−01ATCGA−IS−A3K8−01ATCGA−DX−A1KY−01ATCGA−LI−A67I−01ATCGA−DX−A3U8−01ATCGA−JV−A5VE−01ATCGA−KD−A5QS−01ATCGA−PC−A5DM−01ATCGA−DX−A3UA−01ATCGA−IW−A3M6−01ATCGA−KF−A41W−01ATCGA−MB−A5Y8−01ATCGA−HB−A2OT−01A

TCGA−SARC−RSEM correlation

histologic_diagnosisDedifferentiated liposarcomaLeiomyosarcoma (LMS)MyxofibrosarcomaUndifferentiated Pleomorphic Sarcoma (UPS)NA

Ded

iffer

entia

ted

lipos

arco

ma

Leio

myo

sarc

oma

(LM

S)

Myx

ofib

rosa

rcom

a

Und

iffer

entia

ted

Pleo

mor

phic

Sar

com

a (U

PS)

no_v

alue

norm

al_t

issu

e10

11

12

13

14

15

16

17

TCGA−SARC−RSEM CDK4 histologic_diagnosis

min pval = 2.54443103386852e−19

CD

K4

Ded

iffer

entia

ted

lipos

arco

ma

Leio

myo

sarc

oma

(LM

S)

Myx

ofib

rosa

rcom

a

Und

iffer

entia

ted

Pleo

mor

phic

Sar

com

a (U

PS)

no_v

alue

norm

al_t

issu

e

10

11

12

13

14

15

16

17

TCGA−SARC−RSEM CDK4 histologic_diagnosis

min pval = 2.54443103386852e−19

CD

K4

CDK4 ACVRL1

c)b)

d)

a)

test for significant gene expression associated with meta data features by performing a t-test of a gene’s expression grouped by a sample’s meta data. As an example, a search for meta data with significantly different CDK4 expression in the TCGA sarcoma dataset reveals that the histologic diagnosis type has large differences in CDK4 expression levels (lowest p-val = 2.54e-19) as calculated by a pairwise t-test between all groups (Figure 4c). A box plot of the groups from histologic diagnosis shows that the CDK4 values from DDPLS are higher than other sarcomas (Figure 4d). Additional types of analyses include the identification of differentially expressed genes using t-tests of gene expression between groups defined by a gene’s expression, a gene’s mutation status, or a meta value label. For example, one could ask what genes are differentially expressed between samples with high CDK4 vs. low CDK4, samples with mutated EGFR vs. wild type EGFR, or samples annotated as male vs. female. Conversely, one can search for mutated genes which differentially express the query gene. e.g. which gene(s) mutations have higher or lower expression of EGFR than wild-type.

7. Precision Medicine

To support precision medicine, CELLX can be used to generate responder / non-responder hypotheses from cell line screening data. As a retrospective example, one can analyze the cell line sensitivity profile of Palbociclib, a CDK4/6 inhibitor under development for ER+ breast cancer. Published breast cell line IC50 values for Palbociclib[11] show a range of responses. (Figure 5a) CELLX can associate IC50 values with cell line expression, CNV, and mutation data from data sources such as CCLE. Samples divided into two groups by user defined cutoffs, in this case <1uM for responder cell lines (LOW IC50) and > 1uM for non-responder cell lines (HIGH IC50) can be used to identify genes whose expression is significantly different between responder and non-responder cells by calculating t-tests on the expression of ~20,000 genes and displaying a p-value ranked table (Figure 5b). Hierarchical clustering on the top 100 most significant genes, ordering the samples from low to high IC50, and coloring the samples by intrinsic breast subtype as defined by PAM50[12] shows that luminal B and Her2 subtypes tend to be sensitive to Palbociclib whereas cells of the basal subtype tend to be resistant (Figure 5c). Luminal A cell line subtypes were not represented in the screening set. Additionally, CELLX can dynamically generate a combination CNV / mutation table for genes which meet user defined amplification / deletion thresholds or have annotated mutations. A ranked table of p-values from Fisher’s exact test for all genes with either a CNV or mutation alteration (Table 2) highlights genes potentially associated with compound activity. While individually, the appearance of any one gene is not necessarily significant, together the combined results from the expression, CNV, and mutation associations highlight RB1, CCNE1, and to a lesser extent CDKN2A. Specifically, the expression of RB1 was low in resistant cells whereas CDKN2A and CCNE1 were high in resistant cells. Interestingly, unlike other targeted therapies where the small molecule target is often the biomarker of sensitivity (e.g. EGFR, MET, BRAF) the significant Palbociclib biomarkers represent markers of resistance. RB1 deficiency (CNV deletion, STOP mutations, and low expression) and concomitant high CDKN2A expression[13] are characteristics of the basal or triple negative breast subtype status (Figure 5c). Thus, if most of the RB1 deficient samples

belong to the triple negative subtype, the remaining luminal A/B (ER+/ERBB2+/-) and ERBB2+ segments would be enriched for possible CDK4i responders. In support of this notion, luminal B and Her2 breast subtype cell lines are mostly sensitive to CDK4i (Figure 5c). !CELLX can also confirm if the low RB1 expression found in triple negative breast cell lines also occurs in primary tissues by using the TCGA-BRCA breast invasive carcinoma dataset. CELLX can identify the genes that are most differentially expressed between RB1 high (> 9.5) vs. RB1 low (< 9.5) expressing cells using t-tests. Several of the top 100 ranking genes by p-value are related to cell cycle (RB1, CDKN2A, CCNE1) or DNA replication/repair (RFC2, RFC4, MCM5, MCM7, CDT1, NASP, POLK, POLD1, MUTYH, FANCE). Hierarchical clustering and labeling with the intrinsic subtype via PAM50[12] shows that similar to cell lines, we find that tumors with low RB1 and high CCNE1/CDKN2A expression are often of the basal subtype (Figure 6).

ZR75

30_L

OW

CAM

A1_L

OW

MDA

MB1

34VI

_LO

WH

CC

202_

LOW

UAC

C89

3_LO

WEF

M19

_LO

WEF

M19

2A_L

OW

MDA

MB3

61_L

OW

HC

C15

00_L

OW

HC

C14

19_L

OW

MDA

MB4

15_L

OW

HC

C38

_LO

WUA

CC

812_

LOW

HC

C22

18_L

OW

ZR75

1_LO

WM

DAM

B453

_LO

WT4

7D_L

OW

MC

F7_L

OW

BT20

_LO

WBT

474_

LOW

SKBR

3_LO

WKP

L1_L

OW

HC

C11

43_L

OW

MDA

MB2

31_L

OW

HC

C13

95_L

OW

HS5

78T_

LOW

CAL

51_L

OW

HC

C15

69_H

IGH

HC

C70

_HIG

HH

CC

1187

_HIG

HH

CC

1954

_HIG

HM

DAM

B468

_HIG

HH

CC

1806

_HIG

HM

DAM

B436

_HIG

HD

U44

75_H

IGH

MDA

MB1

57_H

IGH

BT54

9_H

IGH

HC

C19

37_H

IGH

SLC9A2UHRF1BP1LFREM2DEGS2KIAA1324ZG16BTBC1D30PRRT3THSD4LARGERB1DCAF5PER2RUNDC1HERC1PI3CDKN2ACMPK2CARD6LYNIFIT2IFIT3B3GNT5CCDC82FMNL2RGS2HHLA3NUPL2OAFTMEM39BC19orf66RNF19BSPATA24PLAGL1KANK4NRN1RSAD2LAMA4SLC1A3PTGS2CCL5LOC375196SOX8CLDN11PLXNA2FAM89AHOXA1MIR31HGMAML2ICAM1RTP4CASP1CARD16CTSSBTN3A3NLRC5HCP5TAPSAR1PLIN2CEBPDCCNE1TTYH3HLA.FNCOA7STX10CDCA8TOB1CENPWCDC20CSTBTMBIM6CLTCTFF1CLDN3SPDEFGPR160CA12ARL1GNPNAT1CSADFAM177A1RGL2CANT1GNSRAB11ALMCD1RSAD1SPATA20PTPN11DIP2BGNPTABSLC39A9SERF2POC1BTMTC3WHAMMRNF103SPTLC2MON2SEL1L

CCLE PD0332991

LOW <= 999 HIGH > 999

−2 0 2Row Z−Score

050

100

150

200

250

300

Color Keyand Histogram

Cou

nt

PAM50BasalHer2LumBNA

MDA

MB1

75ZR

7530

CAM

A1M

DAM

B134

VIH

CC

202

UAC

C89

3EF

M19

SUM

190

EFM

192A

MDA

MB3

61H

CC

1500

HC

C14

19M

DAM

B415

HC

C38

MC

F10A

UAC

C81

2H

CC

2218

ZR75

1M

DAM

B453

184A

1T4

7DM

CF7

BT20

MDA

MB4

35BT

474

SKBR

3KP

L1H

CC

1143

MDA

MB2

31H

CC

1395

SUM

225C

WN

HS5

78T

184B

5UA

CC

732

CAL

51H

CC

1569

CO

LO82

4H

CC

70H

CC

1187

HC

C19

54M

DAM

B468

HC

C18

06M

DAM

B436

DU

4475

MDA

MB1

57BT

549

HC

C19

37

PD0332991 data

IC50

[nM

]

5

10

20

50

100

200

500

1000IC50<100<500<25002500+

a) c)

Figure 5. a) Waterfall plot of breast cell line responses to Palbociclib (PD0332991) colored by IC50 range. b) Example output listing the p-value of genes. dm = difference in group means, statistic = t-statistic (LOW-HIGH), p.value = uncorrected p-value of two-sided, two-class t-test with equal variances. Not shown: FDR and Hochberg adjusted p-values. c) Heatmap of gene expression of top 100 genes by t-test between sensitive (IC50 < 999nM, LOW) and resistant cell lines (IC50 > 999nM, HIGH). The positions of RB1, CDKN2A, and CCNE1 are denoted with arrows. Cell lines are ordered by IC50 and colored by intrinsic breast subtype via PAM50.

b)

8. Summary

CELLX is an informatics infrastructure to manage multi-dimensional genomics datasets containing expression, copy number variation, mutation, and compound sensitivity information. A browser based web page enables an accessible way to visualize, analyze, and download the database data in a pre-formatted table suitable for offline computation. CELLX is presently

cell_name PD0332991 RESPONSE RB1 PIK3C2G CCNE1 CDKN2Apvalue 0.0004 0.0048 0.0136 0.1362MDAMB175 4 LOWZR7530 5 LOW p.P129delCAMA1 8 LOWMDAMB134VI 13 LOW p.P129delHCC202 21 LOWUACC893 24 LOWEFM19 27 LOW p.0?/-2.16SUM190 28 LOWEFM192A 42 LOWMDAMB361 44 LOW p.P129del p.M52IHCC1500 45 LOW 1.27 -2.24HCC1419 51 LOW p.P129delHCC38 64 LOW p.P129del p.0?/-2.75MDAMB415 64 LOW p.P129delMCF10A 92 LOWUACC812 96 LOW 1.26 p.P129delHCC2218 100 LOW p.P129delZR751 110 LOW p.P129delMDAMB453 115 LOW p.P129del184A1 118 LOWT47D 127 LOW p.P129delMCF7 148 LOW p.0?/-2.19BT20 177 LOW p.I388S p.P129del p.0?/-2.11MDAMB435 201 LOW p.?BT474 240 LOW -1.07SKBR3 300 LOWKPL1 327 LOW -1.97HCC1143 359 LOWMDAMB231 432 LOW p.P129del p.0?/-2.53HCC1395 472 LOW p.0?/-2.03SUM225CWN 503 LOWHS578T 524 LOW p.0?184B5 538 LOWUACC732 744 LOWCAL51 905 LOW p.P129delMDAMB468 1000 HIGH p.?/-1.89MDAMB436 1000 HIGH p.G203fs*9HCC1954 1000 HIGHHCC1937 1000 HIGH p.T738_R775del38DU4475 1000 HIGH p.0?/-1.92HCC1569 1000 HIGH 2.02HCC1187 1000 HIGHBT549 1000 HIGH p.?/-2.22MDAMB157 1000 HIGH 1.01COLO824 1000 HIGH p.?HCC70 1000 HIGH p.N480delHCC1806 1000 HIGH 1.25 p.0?/-2.25

Table 2. Association of mutations / CNV with response to Palbociclib (PD0332991). a) Ranking of genes by p-value for Fisher’s Exact test. b) Breast cell line table of selected alterations. Breast cell lines are labeled LOW (sensitive) or HIGH (resistant) and marked altered or non-altered for mutation or CNV change in each gene. Cell lines are ordered by Palbociclib IC50 value. Genes with CNV values > abs(1) or mutations from CCLE are marked as altered. CNV units are in log2 diploid genomes. (i.e. 1=~ 4 copies) CCLE mutation nomenclature: del = deletion, p.0 = whole gene deletion, ? = unknown change, fs = frameshift, * = STOP codon

GENE pval GENE pvalRB1 0.0004 ATP9B 0.0611PIK3C2G 0.0048 CAPRIN1 0.0611C19orf12 0.0136 CTIF 0.0611CCNE1 0.0136 DNM2 0.0611LOC284395 0.0136 EHF 0.0611PLEKHF1 0.0136 ELP2 0.0611POP4 0.0136 EPG5 0.0611URI1 0.0136 FANCI 0.0611VSTM2B 0.0136 HDLBP 0.0611DOCK3 0.0136 LRP6 0.0611NCOA4 0.0136 MAPK4 0.0611ADRA1A 0.0136 MCPH1 0.0611CTNNA1 0.0136 NKX6.3 0.0611TCF12 0.0136 PDCD6 0.0611CDH1 0.0459 PEBP4 0.0611ANKS1B 0.0459 PTK2B 0.0611DIP2C 0.0459 RP1L1 0.0611GSTT1 0.0595 SGK223 0.0611GSTTP2 0.0595 SMAD4 0.0611LOC391322 0.0595 ZFYVE26 0.0611D2HGDH 0.0611 MTAP 0.0932DHRS4L1 0.0611 USP32 0.0932DHRS4L2 0.0611 BCAS1 0.0932ELAC1 0.0611 TRIM37 0.0932GAL3ST2 0.0611 PIK3CA 0.0952LINC00906 0.0611 TP53 0.0952LINC01029 0.0611 AUTS2 0.0971LOC100420587 0.0611 LOC649352 0.0971LOC100505835 0.0611 MIR4650.1 0.0971LOC102724958 0.0611 MIR4650.2 0.0971LOC439994 0.0611 SIGLEC14 0.0971MIR6511B1 0.0611 FHIT 0.0971NAALADL2 0.0611 PIK3C2B 0.0971NUTM2A.AS1 0.0611 PTEN 0.1176RBFOX1 0.0611 CDKN2A 0.1362SALL3 0.0611 LOC284344 0.1560UGT2B28 0.0611 LPAR6 0.1560UQCRFS1 0.0611 NRG1 0.1560APC 0.0611 PDE4D 0.1560BTK 0.0611 EEF2K 0.1560ELN 0.0611 EPHB3 0.1560EPHB6 0.0611 ITPR1 0.1560GCNT2 0.0611 KIAA1549 0.1560HIPK2 0.0611 MAP3K19 0.1560KLK15 0.0611 MELK 0.1560NOS2 0.0611 MLKL 0.1560OMG 0.0611 MMP8 0.1560TBX22 0.0611 MYLK 0.1560ZNF142 0.0611 PLCB2 0.1560AGPAT5 0.0611 SPTA1 0.1560

a) b)

!focused on supporting oncology precision medicine through the evaluation of preconceived hypotheses as well as unbiased, data driven hypothesis generation. Though usable by the general user, CELLX is aimed at the computational biologist who desires more control over the data or wants to integrate custom data not available in public databases.

9. Data Processing

When available, summarized data from the source was used for TCGA, CCLE, and Tumorscape except for CNV calls. If Affymetrix SNP files were available, they were processed relative to the hg18 assembly using the aroma.affymetrix R package according to the methods of H. Bengtsson et al.[14] using the average baseline of 128 female HapMap samples[15] as the reference to maintain consistency and comparability across datasets. Microarray expression data from GEO, Sanger, and CCLE were GC Robust Multiarray Average normalized using R and the gcrma[16] library. Comparable to the TCGA RNA-Seq RSEM pipeline, CCLE RNA-Seq[17] data was processed using RSEM[8] on RefSeq sequences, quartile normalized to 1000, and log2 transformed. The R library genefu[18] predicted PAM50 subtypes and genefilter[19] enabled fast t-tests, F-tests, and correlations. Plots were made using CELLX and edited using Preview and Pages.

Acknowledgements

The results published here are in whole or part based upon data generated by The Cancer Genome Atlas pilot project established by the NCI and NHGRI. Information about TCGA and the investigators and institutions who constitute the TCGA research network can be found at http://cancergenome.nih.gov/ dbGaP Study Accession: phs000178.v8.p7 We thank Andy Futreal and the Wellcome Trust Sanger Institute for generously providing access to cell molecular profiling data. We also thank Adam Pavlicek and Shibing Deng for help with R and Heather Estrella for discussion and feedback.

C9o

rf53

NEK

5KI

AA08

25C

TU1

CC

NE1

CD

KN2A

CD

T1FA

NC

EPS

RC

1M

UTY

HW

DR

4IM

PA2

HAU

S5R

FC4

UQ

CR

HL

MAG

OH

POLD

1C

CH

CR

1G

TF2H

4M

RPL

2AR

NBE

AC

PEB3

KBTB

D7

FAM

63B

INPP

4BU

GC

GTM

EM14

4PO

LKFA

M11

4A2

ARID

4AFA

M17

9BR

BM43

ARL1

5U

EVLD

CR

EBR

FD

NAJ

B14

KLH

L28

ITM

2BAH

NAK

IL6S

TFA

M19

8BAR

HG

EF12

LRR

FIP1

SEL1

LC

CPG

1SP

G11

AFF1

LIM

A1FB

XL5

CAS

C4

TMED

7PJ

A2N

ASP

MC

M7

KLH

DC

3U

2AF1

DAZA

P1U

QC

RH

MR

PL37

ATP8

B1SM

IM14

FAM

214A

SPO

PLFY

CO

1SN

AP23

TMEM

87B

FBXL

3U

BL3

AKAP

11FN

DC

3AEL

F1ZC

3H13

RR

P1R

FC2

ALYR

EFR

ANBP

1M

CM

5ER

I3SN

RPA

CPE

B2LY

SMD

3FC

HO

2VW

A8C

OG

6N

HLR

C3

RC

BTB1

SUC

LA2

CO

G3

CPE

B4R

B1VP

S36

KIAA

1430

MFA

P3U

TP14

CC

TSO

CD

K17

UBR

1AT

G2B

CO

L4A3

BPST

AM2

TCGA−A2−A3XV−01ATCGA−LL−A5YM−01ATCGA−LL−A7SZ−01ATCGA−OL−A66P−01ATCGA−LL−A6FQ−01ATCGA−E9−A6HE−01ATCGA−OL−A66O−01ATCGA−V7−A7HQ−01ATCGA−E2−A576−01ATCGA−LL−A5YN−01ATCGA−AC−A3QQ−01ATCGA−AC−A62V−01ATCGA−E2−A572−01ATCGA−E9−A3X8−01ATCGA−AC−A3TM−01ATCGA−OL−A66L−01ATCGA−A2−A0CR−01ATCGA−AC−A2QJ−01ATCGA−A7−A56D−01ATCGA−AQ−A54N−01ATCGA−E2−A574−01ATCGA−LL−A6FR−01ATCGA−A2−A3XU−01ATCGA−B6−A0I1−01ATCGA−E2−A1LK−01ATCGA−AN−A04D−01ATCGA−AR−A0TU−01ATCGA−AO−A1KR−01ATCGA−AC−A62X−01ATCGA−B6−A0IQ−01ATCGA−AC−A7VC−01ATCGA−BH−A0E6−01ATCGA−OL−A5S0−01ATCGA−OL−A5D7−01ATCGA−A2−A3Y0−01ATCGA−OL−A5RW−01ATCGA−A7−A6VW−01ATCGA−A2−A04U−01ATCGA−A2−A0D0−01ATCGA−HN−A2NL−01ATCGA−E9−A22G−01ATCGA−GM−A3XL−01ATCGA−A1−A0SK−01ATCGA−E2−A1LG−01ATCGA−A2−A0YJ−01ATCGA−B6−A402−01ATCGA−A7−A13E−01ATCGA−AO−A0J6−01ATCGA−B6−A0IJ−01ATCGA−AN−A0AT−01ATCGA−AR−A5QQ−01ATCGA−AC−A2QH−01ATCGA−E2−A1LL−01ATCGA−BH−A0DL−01ATCGA−BH−A18V−06ATCGA−AO−A0J4−01ATCGA−BH−A18V−01ATCGA−E9−A244−01ATCGA−A2−A0T0−01ATCGA−A8−A08X−01ATCGA−A7−A0DA−01ATCGA−GI−A2C9−01ATCGA−AR−A0U4−01ATCGA−D8−A1XQ−01ATCGA−BH−A0BG−01ATCGA−E2−A150−01ATCGA−A8−A08R−01ATCGA−D8−A142−01ATCGA−AN−A0AR−01ATCGA−OL−A5RZ−01ATCGA−D8−A143−01ATCGA−D8−A1Y3−01ATCGA−A2−A3XW−01ATCGA−D8−A1XL−01ATCGA−E9−A1RG−01ATCGA−A7−A5ZV−01ATCGA−A2−A3XX−01ATCGA−B6−A409−01ATCGA−EW−A6SB−01ATCGA−E9−A5FL−01ATCGA−E2−A1B6−01ATCGA−E2−A1LH−01ATCGA−E9−A1NC−01ATCGA−AC−A2BK−01ATCGA−A1−A0SO−01ATCGA−AR−A0TP−01ATCGA−C8−A1HJ−01ATCGA−A2−A0D2−01ATCGA−B6−A0I6−01ATCGA−AN−A0FJ−01ATCGA−E9−A1N8−01ATCGA−E2−A158−01ATCGA−E2−A14Y−01ATCGA−D8−A1JM−01ATCGA−A8−A07R−01ATCGA−AO−A124−01ATCGA−AC−A6IW−01ATCGA−E2−A14R−01ATCGA−E2−A1LI−01ATCGA−AR−A1AQ−01ATCGA−BH−A5IZ−01ATCGA−OL−A6VO−01ATCGA−B6−A0I2−01ATCGA−OL−A66I−01ATCGA−A2−A04P−01ATCGA−E2−A573−01ATCGA−BH−A18T−01ATCGA−BH−A1F6−01ATCGA−E2−A1IP−01ATCGA−A7−A4SD−01ATCGA−AR−A1AI−01ATCGA−A2−A0YM−01ATCGA−A2−A0CM−01ATCGA−D8−A1XK−01ATCGA−A2−A04T−01ATCGA−C8−A134−01ATCGA−C8−A27B−01ATCGA−EW−A1PB−01ATCGA−A1−A0SP−01ATCGA−C8−A12K−01ATCGA−E2−A1II−01ATCGA−E2−A14N−01ATCGA−A2−A0T2−01ATCGA−EW−A1PH−01ATCGA−EW−A1P8−01ATCGA−AR−A1AH−01ATCGA−BH−A0E0−01ATCGA−A8−A07O−01ATCGA−BH−A0RX−01ATCGA−A7−A13D−01ATCGA−EW−A3U0−01ATCGA−A7−A6VV−01ATCGA−B6−A3ZX−01ATCGA−B6−A0RE−01ATCGA−AR−A1AY−01ATCGA−EW−A1OW−01ATCGA−BH−A0B9−01ATCGA−BH−A1FC−01ATCGA−A2−A3XT−01ATCGA−AO−A129−01ATCGA−BH−A0B3−01ATCGA−GM−A2DF−01ATCGA−A2−A3XS−01ATCGA−B6−A400−01ATCGA−E9−A3QA−01ATCGA−A7−A6VY−01ATCGA−LL−A5YP−01ATCGA−AO−A0JB−01ATCGA−EW−A423−01ATCGA−LL−A7T0−01ATCGA−PE−A5DC−01ATCGA−BH−A42T−01ATCGA−LL−A5YO−01ATCGA−E9−A3HO−01ATCGA−BH−A0HU−01ATCGA−AO−A0J2−01ATCGA−B6−A0RH−01ATCGA−C8−A131−01ATCGA−GM−A2DB−01ATCGA−GM−A2DD−01ATCGA−AN−A0FL−01ATCGA−A2−A4S1−01ATCGA−AO−A0JL−01ATCGA−BH−A18G−01ATCGA−B6−A1KF−01ATCGA−AO−A126−01ATCGA−AQ−A54O−01ATCGA−C8−A12Y−01ATCGA−AO−A03T−01ATCGA−LL−A5YL−01ATCGA−LL−A6FP−01ATCGA−E9−A5FK−01ATCGA−OL−A66N−01ATCGA−BH−A5J0−01ATCGA−LL−A50Y−01ATCGA−AC−A3TN−01ATCGA−BH−A42U−01ATCGA−AR−A2LJ−01ATCGA−LD−A66U−01ATCGA−A2−A0CY−01ATCGA−B6−A0IE−01ATCGA−AN−A04C−01ATCGA−A2−A0D1−01ATCGA−B6−A0IB−01ATCGA−A7−A4SF−01ATCGA−E2−A14P−01ATCGA−AN−A0AL−01ATCGA−AN−A0G0−01ATCGA−AR−A24Q−01ATCGA−AO−A12F−01ATCGA−C8−A130−01ATCGA−A2−A3XZ−01ATCGA−LL−A441−01ATCGA−B6−A0X1−01ATCGA−AR−A256−01ATCGA−A2−A04W−01ATCGA−B6−A0WX−01ATCGA−AR−A0U1−01ATCGA−EW−A1PC−01BTCGA−C8−A1HF−01ATCGA−A8−A08L−01ATCGA−D8−A13Y−01ATCGA−AN−A0FY−01ATCGA−GM−A2DH−01ATCGA−BH−A0EE−01ATCGA−C8−A26Y−01ATCGA−AC−A7VB−01ATCGA−E2−A14X−01ATCGA−AR−A0U0−01ATCGA−A7−A0CE−01ATCGA−EW−A1P4−01ATCGA−A2−A0YE−01ATCGA−AN−A0XU−01ATCGA−D8−A27F−01ATCGA−AR−A0TS−01ATCGA−A7−A4SE−01ATCGA−AR−A2LR−01ATCGA−AN−A0FX−01ATCGA−B6−A0RU−01ATCGA−C8−A12V−01ATCGA−AO−A128−01ATCGA−A8−A07U−01ATCGA−AR−A1AJ−01ATCGA−A2−A3XY−01ATCGA−A2−A04Q−01ATCGA−E9−A1ND−01ATCGA−D8−A1JL−01ATCGA−A8−A07C−01ATCGA−D8−A147−01ATCGA−BH−A0WA−01ATCGA−E2−A159−01ATCGA−BH−A0AV−01ATCGA−E2−A1LS−01ATCGA−AQ−A04L−01BTCGA−AR−A251−01ATCGA−E9−A248−01ATCGA−A2−A1G1−01ATCGA−AR−A1AR−01ATCGA−A7−A26I−01ATCGA−BH−A0BL−01ATCGA−E2−A1AZ−01ATCGA−C8−A12P−01ATCGA−AN−A0AM−01ATCGA−JL−A3YW−01ATCGA−C8−A26X−01ATCGA−A2−A4RX−01ATCGA−EW−A1P7−01ATCGA−AR−A2LH−01ATCGA−D8−A13Z−01ATCGA−BH−A18Q−01ATCGA−A2−A0SX−01ATCGA−A2−A0ST−01ATCGA−AQ−A04J−01ATCGA−B6−A0RT−01ATCGA−GM−A2DC−01ATCGA−AN−A03X−01ATCGA−B6−A0WW−01ATCGA−E9−A5UP−01ATCGA−GM−A3NW−01ATCGA−OL−A66K−01ATCGA−A2−A4S2−01ATCGA−AC−A5XU−01ATCGA−E2−A107−01ATCGA−OL−A66H−01ATCGA−A7−A3RF−01ATCGA−C8−A133−01ATCGA−E2−A1IK−01ATCGA−AO−A12H−01ATCGA−AO−A03N−01BTCGA−C8−A1HK−01ATCGA−A8−A06R−01ATCGA−BH−A0GZ−01ATCGA−AO−A0JM−01ATCGA−A2−A04X−01ATCGA−B6−A408−01ATCGA−A2−A0SW−01ATCGA−AC−A62Y−01ATCGA−E2−A1B5−01ATCGA−C8−A1HI−01ATCGA−EW−A2FS−01ATCGA−GM−A2DK−01ATCGA−B6−A0IN−01ATCGA−AO−A0J9−01ATCGA−C8−A1HL−01ATCGA−BH−A18K−01ATCGA−B6−A0WZ−01ATCGA−EW−A1P3−01ATCGA−GM−A2DL−01ATCGA−BH−A0W7−01ATCGA−E2−A2P6−01ATCGA−GM−A2D9−01ATCGA−GM−A2DA−01ATCGA−A7−A3IY−01ATCGA−E9−A54X−01ATCGA−A2−A04Y−01ATCGA−E9−A22A−01ATCGA−EW−A2FR−01ATCGA−AN−A0FT−01ATCGA−C8−A138−01ATCGA−GM−A3NY−01ATCGA−AC−A5XS−01ATCGA−E2−A570−01ATCGA−E2−A56Z−01ATCGA−AO−A125−01ATCGA−A2−A0CQ−01ATCGA−D8−A1JS−01ATCGA−AO−A12B−01ATCGA−AC−A3YI−01ATCGA−E9−A1R2−01ATCGA−A2−A0CP−01ATCGA−E2−A1B4−01ATCGA−GM−A5PV−01ATCGA−A7−A5ZX−01ATCGA−AC−A3BB−01ATCGA−A7−A426−01ATCGA−OL−A5DA−01ATCGA−EW−A3E8−01BTCGA−BH−A0HN−01ATCGA−E2−A10C−01ATCGA−AC−A3YJ−01ATCGA−B6−A0I5−01ATCGA−BH−A0BT−01ATCGA−E2−A14Z−01ATCGA−EW−A424−01ATCGA−JL−A3YX−01ATCGA−B6−A40C−01ATCGA−A7−A425−01ATCGA−B6−A40B−01ATCGA−PE−A5DD−01ATCGA−E2−A2P5−01ATCGA−B6−A401−01ATCGA−LD−A7W5−01ATCGA−BH−A6R8−01ATCGA−AO−A03R−01ATCGA−A8−A08F−01ATCGA−AO−A1KQ−01ATCGA−E2−A14O−01ATCGA−BH−A0HB−01ATCGA−A2−A0SV−01ATCGA−E2−A14W−01ATCGA−GM−A4E0−01ATCGA−A2−A0CK−01ATCGA−OL−A66J−01ATCGA−GM−A5PX−01ATCGA−E2−A1IF−01ATCGA−AC−A3QP−01ATCGA−GM−A3XG−01ATCGA−C8−A1HG−01ATCGA−E2−A1LE−01ATCGA−A2−A0T1−01ATCGA−A2−A0EQ−01ATCGA−BH−A0AY−01ATCGA−BH−A0B6−01ATCGA−EW−A1IZ−01ATCGA−AC−A3W6−01ATCGA−AC−A3W5−01ATCGA−A2−A0EP−01ATCGA−OK−A5Q2−01ATCGA−B6−A0RL−01ATCGA−B6−A0X5−01ATCGA−A2−A0D4−01ATCGA−E9−A5UO−01ATCGA−A8−A08B−01ATCGA−EW−A1OV−01ATCGA−EW−A1OZ−01ATCGA−C8−A135−01ATCGA−C8−A12X−01ATCGA−C8−A12L−01ATCGA−B6−A0IG−01ATCGA−C8−A12T−01ATCGA−BH−A0DK−01ATCGA−D8−A27N−01ATCGA−E9−A1N9−01ATCGA−A2−A4RW−01ATCGA−E9−A295−01ATCGA−BH−A18J−01ATCGA−C8−A12U−01ATCGA−EW−A6SA−01ATCGA−A2−A4S3−01ATCGA−B6−A0RS−01ATCGA−E9−A247−01ATCGA−E9−A22D−01ATCGA−BH−A209−01ATCGA−A2−A0CW−01ATCGA−A8−A06X−01ATCGA−D8−A27R−01ATCGA−BH−A0BZ−01ATCGA−D8−A1JK−01ATCGA−D8−A1Y2−01ATCGA−AO−A1KT−01ATCGA−BH−A0AW−01ATCGA−OL−A5RY−01ATCGA−LL−A73Z−01ATCGA−A8−A08I−01ATCGA−A8−A079−01ATCGA−S3−A6ZH−01ATCGA−A8−A085−01ATCGA−BH−A0DD−01ATCGA−A7−A13F−01ATCGA−A8−A06O−01ATCGA−E9−A1RB−01ATCGA−A7−A6VX−01ATCGA−EW−A1OY−01ATCGA−BH−A0BR−01ATCGA−BH−A1FN−01ATCGA−E2−A1L7−01ATCGA−D8−A1XJ−01ATCGA−A8−A07Z−01ATCGA−C8−A1HM−01ATCGA−AO−A03O−01ATCGA−A7−A0CJ−01ATCGA−BH−A0C0−01ATCGA−AN−A03Y−01ATCGA−AN−A0AJ−01ATCGA−A8−A092−01ATCGA−LQ−A4E4−01ATCGA−E9−A2JS−01ATCGA−A8−A09E−01ATCGA−A8−A07W−01ATCGA−C8−A12W−01ATCGA−AR−A24H−01ATCGA−AN−A0AK−01ATCGA−BH−A0HY−01ATCGA−AO−A03M−01BTCGA−C8−A26W−01ATCGA−E2−A15S−01ATCGA−A8−A082−01ATCGA−A8−A07L−01ATCGA−BH−A18U−01ATCGA−AN−A0XR−01ATCGA−AR−A1AS−01ATCGA−D8−A1Y1−01ATCGA−E2−A14V−01ATCGA−AC−A3EH−01ATCGA−BH−A2L8−01ATCGA−A1−A0SF−01ATCGA−AR−A1AT−01ATCGA−AC−A6NO−01ATCGA−A2−A0T5−01ATCGA−OL−A5D8−01ATCGA−OL−A5RX−01ATCGA−D8−A3Z6−01ATCGA−PE−A5DE−01ATCGA−GM−A3XN−01ATCGA−E2−A1BD−01ATCGA−B6−A0RM−01ATCGA−OL−A6VR−01ATCGA−D8−A1XZ−01ATCGA−D8−A3Z5−01ATCGA−AC−A3W7−01ATCGA−LD−A7W6−01ATCGA−BH−A42V−01ATCGA−LD−A74U−01ATCGA−BH−A28O−01ATCGA−LL−A440−01ATCGA−A7−A13H−01ATCGA−A2−A4RY−01ATCGA−D8−A4Z1−01ATCGA−AO−A0J3−01ATCGA−E2−A1LA−01ATCGA−B6−A0RG−01ATCGA−BH−A0B4−01ATCGA−A2−A25D−01ATCGA−A7−A4SA−01ATCGA−A8−A09M−01ATCGA−A8−A08P−01ATCGA−E9−A1NI−01ATCGA−A2−A04V−01ATCGA−BH−A0HA−01ATCGA−OL−A5RU−01ATCGA−AR−A5QM−01ATCGA−BH−A0DQ−01ATCGA−AR−A5QP−01ATCGA−AC−A6IV−01ATCGA−AQ−A7U7−01ATCGA−E9−A1NA−01ATCGA−A8−A09W−01ATCGA−A8−A06U−01ATCGA−AC−A3OD−01ATCGA−BH−A0HF−01ATCGA−BH−A0BC−01ATCGA−B6−A0X0−01ATCGA−A8−A08A−01ATCGA−A2−A0CS−01ATCGA−BH−A0B0−01ATCGA−A2−A0CV−01ATCGA−AO−A12G−01ATCGA−BH−A18H−01ATCGA−BH−A8FZ−01ATCGA−BH−A0B2−01ATCGA−BH−A0BQ−01ATCGA−AC−A23G−01ATCGA−B6−A0WY−01ATCGA−AC−A2FE−01ATCGA−AN−A04A−01ATCGA−A8−A09D−01ATCGA−AN−A0FD−01ATCGA−BH−A1FU−01ATCGA−E2−A108−01ATCGA−A8−A08H−01ATCGA−AR−A1AO−01ATCGA−A8−A0A7−01ATCGA−AR−A0TX−01ATCGA−BH−A0B7−01ATCGA−A2−A0T4−01ATCGA−AN−A0XW−01ATCGA−AN−A0XN−01ATCGA−BH−A0AZ−01ATCGA−EW−A1P1−01ATCGA−BH−A0H7−01ATCGA−AN−A049−01ATCGA−E9−A1QZ−01ATCGA−BH−A0BD−01ATCGA−BH−A0DI−01ATCGA−A2−A0T7−01ATCGA−E2−A1IN−01ATCGA−A8−A09T−01ATCGA−A7−A13E−01BTCGA−A7−A0DB−01CTCGA−A7−A26E−01BTCGA−A7−A26J−01BTCGA−A7−A13G−01BTCGA−AC−A3OD−01BTCGA−A7−A0DC−01BTCGA−AC−A3QQ−01BTCGA−D8−A1JF−01ATCGA−BH−A1F5−01ATCGA−BH−A1FM−01ATCGA−AO−A1KS−01ATCGA−D8−A1JA−01ATCGA−BH−A1F8−01ATCGA−D8−A1J8−01ATCGA−EW−A1IY−01ATCGA−D8−A1JT−01ATCGA−A2−A1FV−01ATCGA−AO−A1KP−01ATCGA−E2−A1L8−01ATCGA−EW−A1J3−01ATCGA−E9−A1R7−01ATCGA−AR−A252−01ATCGA−EW−A1OX−01ATCGA−A2−A0YF−01ATCGA−E9−A226−01ATCGA−BH−A1F2−01ATCGA−A2−A1FW−01ATCGA−BH−A1EX−01ATCGA−AC−A23C−01ATCGA−B6−A0RO−01ATCGA−E2−A15I−01ATCGA−D8−A27E−01ATCGA−D8−A1XV−01ATCGA−D8−A1XC−01ATCGA−EW−A1J1−01ATCGA−BH−A204−01ATCGA−A2−A0ET−01ATCGA−A8−A091−01ATCGA−A2−A0ER−01ATCGA−A8−A09V−01ATCGA−A2−A259−01ATCGA−AR−A2LE−01ATCGA−BH−A18N−01ATCGA−E2−A156−01ATCGA−A8−A0A2−01ATCGA−E9−A1R5−01ATCGA−BH−A0E7−01ATCGA−A2−A1G0−01ATCGA−E9−A1R3−01ATCGA−EW−A1PG−01ATCGA−BH−A0BO−01ATCGA−E2−A1BC−01ATCGA−BH−A1ET−01ATCGA−E9−A24A−01ATCGA−BH−A1EU−01ATCGA−A7−A26E−01ATCGA−AR−A24M−01ATCGA−D8−A27V−01ATCGA−E2−A1B1−01ATCGA−AO−A1KO−01ATCGA−AC−A2FK−01ATCGA−AR−A24W−01ATCGA−BH−A1FH−01ATCGA−A2−A0EW−01ATCGA−A2−A0ES−01ATCGA−A2−A0EM−01ATCGA−BH−A0HI−01ATCGA−BH−A0EB−01ATCGA−E2−A15M−01ATCGA−AC−A2FF−01ATCGA−A2−A25A−01ATCGA−AC−A23E−01ATCGA−GM−A2DI−01ATCGA−AR−A2LQ−01ATCGA−A2−A0YK−01ATCGA−A7−A0CG−01ATCGA−D8−A1XY−01ATCGA−AR−A24V−01ATCGA−AQ−A0Y5−01ATCGA−GI−A2C8−01ATCGA−B6−A0RN−01ATCGA−BH−A0DE−01ATCGA−E9−A1NH−01ATCGA−A1−A0SH−01ATCGA−BH−A0EA−01ATCGA−BH−A0DV−01ATCGA−A8−A07J−01ATCGA−AO−A12C−01ATCGA−D8−A1XO−01ATCGA−D8−A1XB−01ATCGA−D8−A27L−01ATCGA−E2−A15C−01ATCGA−EW−A1IX−01ATCGA−AR−A2LN−01ATCGA−A2−A0T6−01ATCGA−BH−A1FB−01ATCGA−EW−A1J2−01ATCGA−E9−A1RD−01ATCGA−AR−A255−01ATCGA−A2−A0YI−01ATCGA−A1−A0SD−01ATCGA−AR−A24T−01ATCGA−E9−A1RI−01ATCGA−D8−A1JU−01ATCGA−BH−A0H3−01ATCGA−A8−A07G−01ATCGA−AO−A0JF−01ATCGA−AR−A24O−01ATCGA−D8−A27P−01ATCGA−D8−A1JH−01ATCGA−AR−A2LM−01ATCGA−AC−A2QH−01BTCGA−A7−A26I−01BTCGA−A7−A26F−01BTCGA−A7−A13D−01BTCGA−EW−A1J6−01ATCGA−D8−A1XW−01ATCGA−B6−A0IA−01ATCGA−BH−A1ES−06ATCGA−AR−A0TV−01ATCGA−LL−A442−01ATCGA−C8−A1HO−01ATCGA−A1−A0SQ−01ATCGA−E2−A106−01ATCGA−E2−A1IG−01ATCGA−C8−A3M8−01ATCGA−D8−A27H−01ATCGA−BH−A0BW−01ATCGA−GM−A2DO−01ATCGA−AO−A0JC−01ATCGA−A1−A0SB−01ATCGA−LL−A73Y−01ATCGA−E9−A243−01ATCGA−E2−A1IH−01ATCGA−D8−A27M−01ATCGA−BH−A1F0−01ATCGA−A2−A25F−01ATCGA−AO−A03U−01BTCGA−BH−A6R9−01ATCGA−BH−A1FE−06ATCGA−BH−A208−01ATCGA−E9−A22E−01ATCGA−E9−A1NF−01ATCGA−EW−A2FW−01ATCGA−AO−A0JD−01ATCGA−B6−A0I9−01ATCGA−AO−A03P−01ATCGA−E2−A14T−01ATCGA−AR−A0U2−01ATCGA−BH−A0HL−01ATCGA−E2−A109−01ATCGA−EW−A1PA−01ATCGA−BH−A203−01ATCGA−E9−A1RH−01ATCGA−AR−A0TY−01ATCGA−D8−A1JG−01BTCGA−E2−A1B0−01ATCGA−A8−A09G−01ATCGA−BH−A1EN−01ATCGA−AN−A0AS−01ATCGA−A8−A07B−01ATCGA−AN−A046−01ATCGA−B6−A0IK−01ATCGA−EW−A6SD−01ATCGA−C8−A137−01ATCGA−A8−A094−01ATCGA−D8−A1JP−01ATCGA−A7−A4SB−01ATCGA−C8−A275−01ATCGA−A1−A0SI−01ATCGA−GM−A2DN−01ATCGA−AC−A5EH−01ATCGA−OL−A5D6−01ATCGA−LL−A740−01ATCGA−E9−A1N6−01ATCGA−C8−A278−01ATCGA−AO−A0JE−01ATCGA−A2−A0CX−01ATCGA−A8−A09X−01ATCGA−A7−A0D9−01ATCGA−B6−A1KN−01ATCGA−C8−A27A−01ATCGA−E2−A105−01ATCGA−C8−A1HN−01ATCGA−AR−A0TQ−01ATCGA−EW−A2FV−01ATCGA−C8−A273−01ATCGA−D8−A1XF−01ATCGA−D8−A1XT−01ATCGA−A7−A26G−01ATCGA−A7−A26F−01ATCGA−BH−A0H9−01ATCGA−D8−A27W−01ATCGA−BH−A18R−01ATCGA−E9−A228−01ATCGA−EW−A1P0−01ATCGA−A7−A3J0−01ATCGA−BH−A1FR−01ATCGA−AO−A0J5−01ATCGA−E9−A1N3−01ATCGA−A2−A4S0−01ATCGA−B6−A0X4−01ATCGA−E9−A245−01ATCGA−C8−A12Z−01ATCGA−AR−A0TW−01ATCGA−B6−A0WV−01ATCGA−D8−A1JC−01ATCGA−A2−A0YH−01ATCGA−AR−A254−01ATCGA−E2−A14S−01ATCGA−S3−A6ZF−01ATCGA−A2−A0YT−01ATCGA−E9−A1R4−01ATCGA−BH−A0E2−01ATCGA−C8−A3M7−01ATCGA−AC−A6IX−01ATCGA−AC−A6IX−06ATCGA−E9−A22B−01ATCGA−A8−A08J−01ATCGA−E9−A227−01ATCGA−E2−A1L9−01ATCGA−A2−A0EN−01ATCGA−AN−A0FN−01ATCGA−E9−A1NG−01ATCGA−A2−A0CZ−01ATCGA−E2−A1IJ−01ATCGA−A7−A5ZW−01ATCGA−C8−A12N−01ATCGA−A2−A04N−01ATCGA−AR−A5QN−01ATCGA−D8−A73U−01ATCGA−AC−A2QI−01ATCGA−A2−A0CO−01ATCGA−AR−A1AM−01ATCGA−BH−A0HK−01ATCGA−BH−A0BP−01ATCGA−BH−A0DG−01ATCGA−BH−A0DX−01ATCGA−E2−A14Q−01ATCGA−B6−A0RQ−01ATCGA−E2−A14U−01ATCGA−AN−A0XS−01ATCGA−A1−A0SG−01ATCGA−EW−A1P6−01ATCGA−MS−A51U−01ATCGA−B6−A0X7−01ATCGA−D8−A73X−01ATCGA−E9−A1R0−01ATCGA−A2−A0EO−01ATCGA−AN−A0FS−01ATCGA−AO−A0J8−01ATCGA−B6−A0IP−01ATCGA−BH−A0H5−01ATCGA−A2−A0SY−01ATCGA−E2−A1IO−01ATCGA−A7−A4SC−01ATCGA−BH−A0H6−01ATCGA−A2−A0EX−01ATCGA−HN−A2OB−01ATCGA−BH−A8G0−01ATCGA−B6−A0IH−01ATCGA−AC−A2FO−01ATCGA−B6−A1KI−01ATCGA−E2−A153−01ATCGA−A8−A06P−01ATCGA−E2−A3DX−01ATCGA−A8−A096−01ATCGA−AR−A1AU−01ATCGA−BH−A0HQ−01ATCGA−D8−A141−01ATCGA−A8−A093−01ATCGA−B6−A0RI−01ATCGA−OL−A5RV−01ATCGA−A7−A3J1−01ATCGA−E9−A1N5−01ATCGA−AR−A2LO−01ATCGA−A2−A3KC−01ATCGA−AC−A3HN−01ATCGA−BH−A0E9−01BTCGA−BH−A0BJ−01ATCGA−D8−A146−01ATCGA−D8−A145−01ATCGA−A2−A0YD−01ATCGA−AC−A2B8−01ATCGA−BH−A0W5−01ATCGA−A2−A0YL−01ATCGA−A8−A08Z−01ATCGA−D8−A1XM−01ATCGA−A2−A1FZ−01ATCGA−BH−A18M−01ATCGA−E2−A10F−01ATCGA−BH−A201−01ATCGA−B6−A0WT−01ATCGA−E9−A229−01ATCGA−B6−A0I8−01ATCGA−AQ−A1H2−01ATCGA−D8−A1JI−01ATCGA−D8−A1JN−01ATCGA−BH−A0BM−01ATCGA−A2−A1G6−01ATCGA−C8−A12M−01ATCGA−AC−A23H−01ATCGA−AN−A0FZ−01ATCGA−AN−A0FW−01ATCGA−AR−A250−01ATCGA−AN−A041−01ATCGA−AO−A0J7−01ATCGA−A8−A06Z−01ATCGA−A8−A06Q−01ATCGA−A2−A0SU−01ATCGA−A2−A25E−01ATCGA−E2−A1IE−01ATCGA−D8−A1X6−01ATCGA−E9−A249−01ATCGA−E9−A1RA−01ATCGA−A8−A09I−01ATCGA−E2−A15A−06ATCGA−E2−A15A−01ATCGA−AN−A0FV−01ATCGA−E2−A1LB−01ATCGA−BH−A0DZ−01ATCGA−A2−A0EY−01ATCGA−AR−A0U3−01ATCGA−E2−A154−01ATCGA−BH−A0W3−01ATCGA−AN−A0FK−01ATCGA−BH−A18L−01ATCGA−BH−A0BA−01ATCGA−A7−A0DB−01ATCGA−BH−A0HP−01ATCGA−BH−A1EY−01ATCGA−C8−A12Q−01ATCGA−D8−A1J9−01ATCGA−BH−A1EV−01ATCGA−A8−A07I−01ATCGA−EW−A6S9−01ATCGA−A8−A076−01ATCGA−A8−A081−01ATCGA−A7−A2KD−01ATCGA−AO−A03L−01ATCGA−S3−A6ZG−01ATCGA−A8−A095−01ATCGA−AN−A0XO−01ATCGA−A8−A086−01ATCGA−A8−A09B−01ATCGA−AR−A1AK−01ATCGA−D8−A27I−01ATCGA−A8−A0A4−01ATCGA−A8−A08S−01ATCGA−EW−A1IW−01ATCGA−AN−A0XL−01ATCGA−B6−A0RV−01ATCGA−A1−A0SN−01ATCGA−D8−A1JJ−01ATCGA−AC−A2FM−01ATCGA−AR−A1AW−01ATCGA−E9−A2JT−01ATCGA−AR−A0TT−01ATCGA−BH−A0BF−01ATCGA−A2−A0CL−01ATCGA−AO−A12D−01ATCGA−AR−A24U−01ATCGA−BH−A0GY−01ATCGA−A8−A075−01ATCGA−E9−A1RF−01ATCGA−BH−A0DS−01ATCGA−A1−A0SM−01ATCGA−B6−A0IO−01ATCGA−AO−A03V−01ATCGA−E2−A15H−01ATCGA−E2−A15K−01ATCGA−A8−A084−01ATCGA−A8−A09C−01ATCGA−AR−A1AN−01ATCGA−BH−A0E1−01ATCGA−A8−A07E−01ATCGA−AR−A24S−01ATCGA−D8−A27G−01ATCGA−BH−A0C3−01ATCGA−D8−A1XS−01ATCGA−AR−A0TZ−01ATCGA−B6−A0IM−01ATCGA−BH−A1EW−01ATCGA−D8−A1Y0−01ATCGA−E9−A22H−01ATCGA−A8−A07P−01ATCGA−AO−A12E−01ATCGA−EW−A1PF−01ATCGA−A2−A25B−01ATCGA−E2−A10A−01ATCGA−AR−A24R−01ATCGA−A7−A26H−01ATCGA−AR−A24P−01ATCGA−A2−A25C−01ATCGA−E9−A1N4−01ATCGA−AR−A1AV−01ATCGA−E2−A10E−01ATCGA−AO−A0JA−01ATCGA−E2−A15T−01ATCGA−A2−A1G4−01ATCGA−A2−A3KD−01ATCGA−D8−A1X7−01ATCGA−E9−A1RE−01ATCGA−A8−A07F−01ATCGA−C8−A274−01ATCGA−BH−A1FE−01ATCGA−E2−A15E−01ATCGA−E2−A15E−06ATCGA−BH−A1FL−01ATCGA−E2−A15G−01ATCGA−A2−A0YC−01ATCGA−EW−A1J5−01ATCGA−AC−A2FB−01ATCGA−BH−A0W4−01ATCGA−A8−A099−01ATCGA−D8−A1JE−01ATCGA−D8−A1XU−01ATCGA−E2−A10B−01ATCGA−A8−A08C−01ATCGA−BH−A0BS−01ATCGA−A8−A07S−01ATCGA−E2−A15J−01ATCGA−A2−A0D3−01ATCGA−D8−A73W−01ATCGA−D8−A1XG−01ATCGA−A8−A08T−01ATCGA−E9−A3Q9−01ATCGA−GM−A2DM−01ATCGA−D8−A1JB−01ATCGA−A8−A0A6−01ATCGA−AR−A24X−01ATCGA−E2−A15D−01ATCGA−BH−A0DT−01ATCGA−AO−A12A−01ATCGA−BH−A0HO−01ATCGA−E2−A15K−06ATCGA−A8−A083−01ATCGA−B6−A0IC−01ATCGA−BH−A28Q−01ATCGA−D8−A27K−01ATCGA−AQ−A1H3−01ATCGA−AR−A1AL−01ATCGA−A7−A0CH−01ATCGA−A7−A13G−01ATCGA−AC−A2FG−01ATCGA−BH−A0DO−01BTCGA−A8−A0A1−01ATCGA−AN−A0XP−01ATCGA−EW−A1PE−01ATCGA−E2−A1L6−01ATCGA−E2−A15L−01ATCGA−A7−A26J−01ATCGA−C8−A1HE−01ATCGA−E2−A1IU−01ATCGA−D8−A1X8−01ATCGA−A8−A09Z−01ATCGA−A8−A09K−01ATCGA−E2−A155−01ATCGA−A8−A06T−01ATCGA−AC−A2BM−01ATCGA−B6−A1KC−01BTCGA−E2−A15R−01ATCGA−AO−A0JI−01ATCGA−A8−A09Q−01ATCGA−AN−A0XT−01ATCGA−BH−A1FD−01ATCGA−BH−A0B8−01ATCGA−C8−A12O−01ATCGA−BH−A202−01ATCGA−A7−A0DC−01ATCGA−A8−A0AB−01ATCGA−BH−A0H0−01ATCGA−A2−A04R−01ATCGA−E2−A15O−01ATCGA−A2−A0CT−01ATCGA−B6−A2IU−01ATCGA−BH−A0DH−01ATCGA−BH−A0BV−01ATCGA−B6−A0WS−01ATCGA−AO−A0JJ−01ATCGA−BH−A0B1−01ATCGA−E2−A152−01ATCGA−D8−A1JD−01ATCGA−A8−A09N−01ATCGA−A2−A0EV−01ATCGA−EW−A1P5−01ATCGA−BH−A0C7−01BTCGA−D8−A1XD−01ATCGA−EW−A1PD−01ATCGA−A2−A0T3−01ATCGA−BH−A1ES−01ATCGA−AR−A1AP−01ATCGA−A1−A0SJ−01ATCGA−BH−A0HX−01ATCGA−BH−A0DP−01ATCGA−D8−A1X5−01ATCGA−AR−A2LK−01ATCGA−A8−A0A9−01ATCGA−AR−A2LL−01ATCGA−BH−A0HW−01ATCGA−A2−A1FX−01ATCGA−A8−A08G−01ATCGA−EW−A6SC−01ATCGA−AN−A0FF−01ATCGA−A2−A0YG−01ATCGA−BH−A18P−01ATCGA−BH−A0B5−01ATCGA−BH−A0C1−01BTCGA−BH−A18F−01ATCGA−A1−A0SE−01ATCGA−A8−A097−01ATCGA−C8−A26Z−01ATCGA−C8−A132−01ATCGA−AR−A1AX−01ATCGA−D8−A27T−01ATCGA−BH−A18I−01ATCGA−E9−A1RC−01ATCGA−AR−A24K−01ATCGA−AR−A24N−01ATCGA−D8−A1XA−01ATCGA−A7−A3IZ−01ATCGA−A7−A0CD−01ATCGA−A8−A0AD−01ATCGA−A8−A090−01ATCGA−C8−A26V−01ATCGA−BH−A1FJ−01ATCGA−AQ−A04H−01BTCGA−A2−A0CU−01ATCGA−BH−A1FG−01ATCGA−AR−A24Z−01ATCGA−A8−A06Y−01ATCGA−E9−A1R6−01ATCGA−A2−A0EU−01ATCGA−BH−A0AU−01ATCGA−E2−A15F−01ATCGA−A8−A09R−01ATCGA−AR−A24L−01ATCGA−BH−A1EO−01ATCGA−BH−A0EI−01ATCGA−A8−A09A−01ATCGA−D8−A1X9−01ATCGA−E9−A1NE−01ATCGA−A8−A08O−01ATCGA−A8−A06N−01ATCGA−D8−A1XR−01ATCGA−AR−A0TR−01ATCGA−AO−A0JG−01ATCGA−D8−A140−01ATCGA−E2−A1IL−01ATCGA−E2−A15P−01ATCGA−AN−A0XV−01ATCGA−B6−A0RP−01ATCGA−BH−A18S−01A

TCGA−BRCA−RSEM

−5 0 5Column Z−Score

020

0040

0060

0080

0010

000

1200

014

000

Color Keyand Histogram

Cou

nt

PAM50.RNASEQBasalHer2LumALumBNA

CCNE1 CDKN2A

RB1

Figure 6. TCGA Breast differential gene expression between RB1 high and RB1 low expressing tumors. Hierarchical clustering of the top 100 genes in a heat map colored by breast subtype as determined by PAM50. Positions of CDKN2A, CCNE1, and RB1 are denoted by arrows.

References !1. Barretina J, et.al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. 2012 Mar 28;483(7391):603-7. PMID: 22460905 2. Yang W, et.al. Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells. Nucleic Acids Res. 2013 Jan;41:D955-61 PMID: 23180760 3. Chin L, Hahn WC, Getz G, Meyerson M. Making sense of cancer genomic data. Genes Dev. 2011 Mar 15;25(6):534-55. doi: 10.1101/gad.2017311. PMID:21406553 4. Wu C, Macleod I, Su AI. BioGPS and MyGene.info: organizing online, gene-centric information. Nucleic Acids Res. 2013 Jan;41:D561-5 PMID: 23175613 5. Beroukhim R, et.al. The landscape of somatic copy-number alteration across human cancers. Nature. 2010 Feb 18;463(7283):899-905. PMID: 20164920 6. Forbes SA, et.al. COSMIC: mining complete cancer genomes in the Catalogue of Somatic Mutations in Cancer. Nucleic Acids Res. 2011 Jan;39:D945-50. PMID: 20952405 7. Cerami E, et.al. The cBio Cancer Genomics Portal: An Open Platform for Exploring Multidimensional Cancer Genomics Data. Cancer Discov. 2012 May;2(5):401-4. PMID: 22588877 8. Li B, Dewey CN. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics. 2011 Aug 4;12:323. doi: 10.1186/1471-2105-12-323. PMID:21816040 9. Baneth V, Raica M, Cîmpean AM. Rom J Assessment of angiogenesis in soft-tissue tumors. Morphol Embryol. 2005;46(4):323-7. PMID:16688371 10. Binh MB, Sastre-Garau X, Guillou L, de Pinieux G, Terrier P, Lagacé R, Aurias A, Hostein I, Coindre JM. MDM2 and CDK4 immunostainings are useful adjuncts in diagnosing well-differentiated and dedifferentiated liposarcoma subtypes: a comparative analysis of 559 soft tissue neoplasms with genetic data. Am J Surg Pathol. 2005 Oct;29(10):1340-7. PMID:16160477 11. Finn RS, Dering J, Conklin D, Kalous O, Cohen DJ, Desai AJ, Ginther C, Atefi M, Chen I, Fowst C, Los G, Slamon DJ. PD 0332991, a selective cyclin D kinase 4/6 inhibitor, preferentially inhibits proliferation of luminal estrogen receptor-positive human breast cancer cell lines in vitro. Breast Cancer Res. 2009;11(5):R77. PMID:19874578 12. Parker JS, Mullins M, Cheang MC, Leung S, Voduc D, Vickery T, Davies S, Fauron C, He X, Hu Z, Quackenbush JF, Stijleman IJ, Palazzo J, Marron JS, Nobel AB, Mardis E, Nielsen TO, Ellis MJ, Perou CM, Bernard PS. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol. 2009 Mar 10;27(8):1160-7. doi: 10.1200/JCO.2008.18.1370. Epub 2009 Feb 9. PMID:19204204 13. Knudsen ES, Knudsen KE. Tailoring to RB: tumour suppressor status and therapeutic response. Nat Rev Cancer. 2008 Sep;8(9):714-24. doi: 10.1038/nrc2401 PMID:19143056 14. Bengtsson H, Irizarry R, Carvalho B, Speed TP (2008) Estimation and assessment of raw copy numbers at the single locus level. Bioinformatics 24: 759– 767. PMID:18204055 15. HapMap data available from http://hapmap.ncbi.nlm.nih.gov/downloads/raw_data/hapmap3_affy6.0/ Originally obtained from Affymetrix, but no longer available from that source. 16. Wu J and Gentry RIwcfJMJ. gcrma: Background Adjustment Using Sequence Information. R package version 2.36.0. http://www.bioconductor.org/packages/release/bioc/html/gcrma.html 17. CCLE RNA-Seq data obtained from The Cancer Genomics Hub (CGHub) https://cghub.ucsc.edu/ 18. Haibe-Kains B, Schroeder M, Bontempi G, Sotiriou C and Quackenbush J (2014). genefu: Relevant Functions for Gene Expression Analysis, Especially in Breast Cancer.. R package version 1.14.0, http://compbio.dfci.harvard.edu. 19. Gentleman R, Carey V, Huber W and Hahne F. genefilter: genefilter: methods for filtering genes from microarray experiments. R package version 1.46.1. http://www.bioconductor.org/packages/release/bioc/html/genefilter.html