Modeling Endocrine Resistance in Breast Cancer Robert Clarke, Ph.D., D.Sc. Professor of Oncology Director, Center for Cancer Systems Biology Dean for Research.

COMPREHENSIVECANCER CENTER at GEORGETOWN UNIVERSITY

Lombardi

Modeling Endocrine Resistance in Breast Cancer

Robert Clarke, Ph.D., D.Sc.Professor of Oncology

Director, Center for Cancer Systems BiologyDean for Research

Georgetown University Medical Center


Lombardi

Robert Clarke, Ph.D., D.Sc.

Systems Biology in Cancer Research

Study of an organism viewed as an integrated and interacting network of genes, proteins, and biochemical reactions that give rise to life…*

*Lee Hood - Institute for Systems Biology

A systems biology approach is required to integrate knowledge from cancer biology with computational and mathematical modeling

● Systems biology goals– interactions among the components of a biological system– how these interactions control system function and behavior– integrate and analyze complex data from multiple sources using

interdisciplinary tools– build in silico models of system (network) function

Systems Biology Research CycleEndocrinologist 94: 13, 2010

Biological cycle

Integration with modeling


Lombardi


Resistance to Endocrine Therapies

To understand how some ER+ breast cancers become (or already are at diagnosis) resistant to endocrine therapies, we invoke an integrated, multimodal, network hypothesis

– network is modular and exhibits both redundancy and degeneracy– signaling is highly integrated and coordinates many cellular functions

In the face of the stresses induced by endocrine therapies, the network modules of interest are those that regulate cell fate, i.e., the cell’s choice:

– to live or die (e.g. control/execution of apoptosis, autophagy, necrosis)– if to live, whether or not to proliferate (i.e., cell cycle control/execution)

Age (Menopausal Status) Risk Reduction1

Recurrence: <50 years (ER+) 45 ± 8%Recurrence: 60-69 years (ER+) 54 ± 5%Recurrence (ER-) 6 ± 11% (not

significant)

Death: any cause <50 years (ER+) 33 ± 6% Death: any cause 60-69 years (ER+) 32 ± 10%Death: any cause (ER-) -3 ± 11% (not

significant)

}Benefit from TAM

1Proportional reduction in the 10-year risk of recurrences or death from the Early Breast Cancer Trialists Group meta analyses


Lombardi

● Compare failures “on-treatment” (early; ≤3yrs) with those that recurred (distant recurrence) later “off-treatment” (later; ≥5 yrs)

● Construct molecular classifiers using gene expression microarray

data from breast tumors collected at diagnosis– integrated resampling workflow to ease the “gene selection bias” problem

– Support Vector Machine with recursive feature elimination

Are all Tamoxifen Failures the Same?

Computational Modeling: task = classification

health care cost (treatment)

ER+

human cost (mortality)

ER-


Lombardi

All S

amples

Split in 10 equal parts

ValidationSet

TrainingSet

TestSet

LearningSet

Split in 1/3 and 2/3 parts

LOOCV to optimal number of genes

Optimal gene set

Validation gene list

ValidationPerformance

Com

bination of 10 validation gene lists

Average Validation Performance

Gene list= 100% occurrence

Repeat 100 times

Classifier

10-fd CV

Repeat 100 times

a:

b:

c:

Classifying Early vs. Later TAM Recurrences

· Resampling approaches used to ease the “gene selection bias” problem

– training procedure (block a)– validation step (block b)

· Must outperform random gene sets of the same size (10,000 random sets)1

· Must meet n=7 pre-established performance benchmarks2

· Clinical characteristics– n=131 cases; >95% ER+; almost all Invasive Ductal Carcinomas– Tamoxifen only after surgery and radiotherapy– ≥15 years of clinical follow-up1Venet et al., PLoS Comp Biol, 2011 report that >60% (in some cases up to 90%) of breast cancer molecular predictors are no better than random gene sets

2Mackay et al., JNCI, 2011 report that the molecular subgroup classifications for the LumA, LumB, and normal-like breast cancer subgroups are not robust


Lombardi

Early (≤3 yr) and Later (≥5 yr) TAM Recurrences

Accuracy Specificity Sensitivity AUC PPV NPV Hazard Ratio P-value

0.90 0.95 0.81 0.87 0.91 0.89 3.45 <0.0001

BC030280

Loi et al.Accuracy Specificity Sensitivity AUC PPV NPV Hazard Ratio P-value

0.77 0.83 0.74 0.81 0.88 0.67 3.11 0.0004

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Se

nsi

tivity

1 - Specificity

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

THSD4THSD4ERBB4

SLC7A8RBM24KIAA1467CMYA5STK32BMAOBNCOA7MUM1L1DEGS2FLJ14959EFHC2CX3CR1PTGER3PTGER3PTGER3ITGA8MS4A7MS4A7SEC14L2FERD3LTNNI1C1orf86STK35RNF133ZNF704MGC52498LOC283079KCNJ12LOC651964

RHDC8orf12OFCC1

SLC6A6NAP1L4GGNOR10A3PRO0471C12orf65LOC440292 /// LOC647995LOC150763DCLK3IKZF1LOC284801CR1LTMEM4BATF2LRP8SOD2C1orf187SLC7A5PNPLA3ME3ATXN7L1C1orf96LOC144874PLCH1ADAMTS1BCL2L14USP36RFX3LOC728683TAAR3STXBP5LRAB6BRASD2

%

Su

rviv

al

Time

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1-specificity

sens

itivi

ty

1 - Specificity

Se

nsi

tivity

%

Su

rviv

al

Time

Performance exceeds all (n=7) pre-established benchmarks in both datasets (and outperforms all of 10,000 randomly selected gene sets)Minetta Liu (Georgetown; Mayo)Mike Dixon; Bill Miller (Edinburgh)Jason Xuan (Virginia Tech)Joseph Wang (Virginia Tech)


Lombardi

Approach to Network Modeling

● We take a systems biology approach to integrate knowledge from cancer biology

with computational and mathematical modeling to make both qualitative and quantitative predictions on how a system functions

● We apply both computational and mathematical modeling tools– computational models can find local topologies or modules within high

dimensional data using multiple different methods (top down)

– mathematical models can represent local topologies or modules by a series of differential equations, stochastic reaction networks, etc. (bottom up)

Chen et al. Nucl Acid Res, in press, 2013 Wang et al., J Mach Learn Res, in press, 2013 Yu et al, Bioinformatics, in revision, 2013Gusev et al., Cancer Informatics, 12: 31-51, 2013 Gu et al. Bioinformatics, 28: 1990-1997, 2012 Tyson et al., Nature Rev Cancer, 11: 523-532, 2011Zhang et al., PLoS ONE, 5 (4): e10268, 2010 Yu et al., J Mach Learn Res, 11;2141-2167, 2010 Chen et al., Bioinformatics, 26: 1426-1422, 2010Zhang et al., Bioinformatics 25: 526-532, 2009 Clarke et al., Nature Rev Cancer 8: 37-49, 2008 Wang et al., Bioinformatics, 23: 2024-2027, 2007

Computational modeling Physical modeling

● The module(s) of interest exist within an immense search space (the human interactome) and we don’t know all of the genes/proteins in each module

● Networks are high dimensional and so the data have unique properties, e.g.,

curse

of dimensionality; confound of multimodality; scale free; small world; etc.Clarke et al., Nature Rev Cancer, 2008; Wang et al., Br J Cancer 2008


Lombardi

Network Modeling: Where to Start?● We have selected our key modules of interest

– live or die (e.g., apoptosis, autophagy, necrosis)– proliferate or growth arrest (i.e., cell cycling)

● We know that ERα is relevant and will coordinate several cell functions– key regulator in normal mammary gland development and function1

– most tumors acquiring endocrine resistance retain ERα expression2

– responses to 2nd and 3rd line endocrine therapies are relatively common2

– small molecule inhibitors and RNAi against ERα inhibit resistant cells3

● We don’t know precisely how ERα signaling is regulated or wired

● We need an ERα-driven network model to guide our studies

1Johnson et al., Nat Med , 20032Clarke et al. Pharmacol Rev, 2002

3Kuske et al., Endocr Relat Cancer, 2006 Wang et al., Cancer Cell, 2006

ERα


Lombardi

Roadmap for Modeling ER-Related Signaling

John Tyson et al., Nature Rev Cancer, 2011

Primary Inputs/Regulators

Estrogen Receptors Growth Factor Receptors (e.g., EGFR; Her2)

Hypothesis: With the stresses induced by endocrine therapies, the network modules of interest are those that regulate cell fate, i.e., the breast cancer cell’s choice – to live or die (e.g. control/execute apoptosis, autophagy, necrosis, UPR)– if to live, whether or not to proliferate (i.e., cell cycle control/execution)

Primary Outputs


Lombardi

· ER is the most upstream regulator of cell fate decisions

· ER can be mutated, phosphorylated, degraded, recycled

– mutations appear to be relatively rare in clinical samples– Fulvestrant acts by targeting the receptor for ubiquitin-mediated

degradation

· ER can activated by ligand or by growth factors– several growth factors and their receptors signaling to MAPKs that

can activate ER through phosphorylation

· Regulation of ER activation may be a central determinant of endocrine responsiveness

ERα as a “Master” Regulator of Cell Fate


Lombardi

ER and EGFR/HER2 Crosstalk

Parameters Description Value

γEPI Rate of EPI reaching its steady state 3×10−4 min-1

γGFR Rate of GFR reaching its steady state5×10−2 min-1

ωEPI Basal inhibition of EPI −1.92

ωGFR Basal inhibition of GFR −4

ωE2ER Basal inhibition of E2ER −2.1

ωERP Basal inhibition of ERP −1.5

ωEPI,GFR EPI activation by GFR 6

ωGFR,GFR GFR activation by EPI 5

ωGFR,E2ER GFR inhibition by E2ER −2

ωGFR,ERP GFR activation by ERP 1.85

ωGFR,GFRover GFR activation by GFRover 0.15

ωE2ER,E2 E2ER activation by E2 3

ωERP,GFR GFR activation by ERP 3

E2 E2 level in MCF7 cells 1 (normal); 0 (E2-depleted cells)

ERT Parameter determining total ER level in MCF7 cells

1 (normal); >1 (ER-overexpressed cells)

GFRover Excess GFR in GFR-transfected MCF7 cells 0 (normal); >0 (GFR-transfected cells)

Crosstalk between ER and GFR

GFR = growth factor receptor (HER2 or EGFR)GFRover = transfected with GFREPI = epigenetic componentsERP = estrogen-independentE2ER= estrogen-dependentERT = total ER levels

Primary data from multiple clones of MCF-7 cellstransfected with either HER2 or EGFR and assayedfor E2-dependent or E2-independent growth

Liu et al., Breast Cancer Res Treat, 1995 Miller et al., Cell Growth Diff, 1994

Chun Chen et al., FEBS Lett, 2013

Mathematical Modeling: task = nature of ER regulation of cell fate


Lombardi

ER is a Bistable Switch for EGFR/HER2 Crosstalk

John Tyson et al., Nature Rev Cancer, 2011Bistability: resting in two different minimum states separated by a maximum

● Breast cancer cells can switch reversibly and robustly between E2

and GFR dependence– GFR can inhibit ER expression and/or activate (phosphorylate) any

remaining ER– cells can eliminate or silence GFR plasmid (epigenetic) and upregulate ER

● Model can explain some of the molecular heterogeneity in cell populations

● Blocking either pathway increases the likelihood that the otherpathway will be activated

● E2-dependence GFR-dependence (ER-independence) occurs more easily/rapidly than the reverse

Mathematical Modeling: task = nature of ER regulation of cell fate

Chun Chen et al., FEBS Lett, 2013


Lombardi


Minimum Action Paths characterize state transitions Intermittent therapy opens a 2nd response window

Shifting E2 dose response

LigandDependent

LigandSupersensitive

LigandIndependent

Phenotype Transitions Support Intermittent Therapy

Chun Chen et al., in preparation

Mathematical Modeling: task = ER-driven phenotype transitions


Lombardi


● What molecular events are associated with endocrine resistance?

● When are these changes acquired (early, late)?

● Which changes are functionally/mechanistically important?

● How do cells coordinate their functions to make and execute a cell fate decision?

Factors Affecting Endocrine Responsiveness


Lombardi

ERα Signaling: Early vs. Late Recurrences

● Identify closest protein partners to ERα using a novel Random Walk (RW) based algorithm with Metropolis Sampling (MS; Markov Chain-Monte Carlo) technique to walk 8 PPI (protein-protein interaction) databases

– 2-steps per iteration (walk)– 300,000 iterations– 1,452 neighbors selected; n=50 are frequently visited

● Model the n=50 using the microarray data and MS/RW method

Minetta Liu et al., in reviewBai Zhang et al., in preparation

Num

ber o

f nod

es

Computational Modeling: task = network topology

Genes Gene Ontology p-value

23/50 “Apoptosis” 2.9E-13

14/50 “Cell proliferation” 6.8E-5

Minetta Liu (Georgetown; Mayo)Mike Dixon; Bill Miller (Edinburgh)Jason Xuan (Virginia Tech)Joseph Wang (Virginia Tech)

Circles = nodes

Lines = edges

NFκB

BCL2

AKT

MAPK

EGFR

ERα

red = overexpressed in ‘Early’

green = overexpressed in ‘Late’

SRC

ERβ

AR

Circles = nodes

Lines = edges

red = overexpressed in ‘Early’

green = overexpressed in ‘Late’

yellow = inconsistent MAPKERα

SRC

ERβ

AR

BCL2

EGFR


Lombardi

· Represent the local structures of a network by a set of local conditional

probability distributions – decompose the entire expression profile into a series of local networks (nodes; parents)

– local dependency is learned– local conditional probabilities are estimated from linear regression model– allow more than one conditional probability distribution per node– Lasso technique is used to limit overfitting

· Identify motifs and “hot spots” within motifs– time series data from T47D cells ± E2; ± Fulvestrant (Lin et al., Genome Biol, 2004)

– key nodes identified include AKT, XBP1, NFκB, several BCL2 family members, several MAPKs

Yue Wang et al., Bioinformatics, 2009

Computational Modeling: Differential Dependency Network (DDN) analysis

Some Changes are Acquired Early

XBP1 is a key component of the Unfolded Protein Response (UPR)

plasma membrane

cytosol

nucleus

extracellularly exposed

plasma membrane

cytosol

nucleus

extracellularly exposed

BCL2 (large family) regulate apoptosis/survival


Lombardi

Gene Name Gene Symbol1 Difference

p-value

Genes Up-regulated in LCC9 vs. LCC1

Cathepsin D CTSD 5-fold <0.001

X-box Binding Protein-1 (TF) XBP1 4-fold <0.001

B-cell CLL/lymphoma 2 BCL2 4-fold <0.001

Epidermal growth factor receptor

EGFR 2-fold 0.002

Heat Shock Protein 27 HSBP1 2-fold 0.001

NFκB (p65) (TF) RELA 2-fold <0.05

Genes Down-regulated in LCC9 vs. LCC1

Death Associated Protein 6 DAXX 6-fold 0.049

Early Growth Response-1 (TF) EGR1 3-fold <0.05

Interferon Regulatory Factor-1 (TF) IRF1 2-fold <0.05

Tumor Necrosis Factor-α TNF 2-fold <0.05

TNF-Receptor 1 TNFRSF1A 2-fold <0.05Data are mean values of the relative level of expression for each gene to the nearest integer; 1HUGO Gene Symbols

UPR = Unfolded Protein Response; TF = transcription factor

Selected from molecular comparison of sensitive (LCC1) vs. stable resistant variant (LCC9)

autophagy

UPR

UPR

apoptosis

apoptosis

apoptosis

Some Early Changes are Retained

Zhiping Gu et al., Cancer Res, 2002Todd Skaar et al, J Steroid Biochem Mol Biol, 1998

apoptosis/UPR

apoptosis

UPR/apoptosis


Lombardi

Symbol Gene NameChang

e

p-value # CREs

APBB2 amyloid beta (A4) precursor protein-binding -1.3 0.001 1

BCL2 B-cell CLL/lymphoma-2 3.1 0.029 3

CRK v-crk sarcoma virus CT10 oncogene homolog -2.0 0.003 2

ESR1 estrogen receptor alpha (ERα) 2.8 0.040 0*

IL24 interleukin 24 -9.7 <0.001 1

MYC v-myc myelocytomatosis viral oncogene homolog 1.6 0.04 1

PHLDA2pleckstrin homology-like domain, family A,

member 2 -3.3 0.004 2

S100A6 S100 calcium binding protein A6 (calcyclin) 2.3 0.001 1

XRCC6 X-ray repair complementing defective repair 6 1.6 0.016 1

XBP1(s) May Control Some Retained Changes

*several ATF6 sites that may be regulated by ATF6:XBP1 heterodimers

Bianca Gomez et al., FASEB J, 2007


Lombardi

Some Retained Changes are Functionally Important

T47D/ XBP1

T47D/ c

T47D/ XBP1T47D/ XBP1

T47D/ cT47D/ c

XBP1(s) confers Antiestrogen Resistance XBP1 cDNA increases BCL2

EtOH TAM FAS

Rela

tive

Bcl-2

:act

in ra

tio

0

1

2

3

4MCF7/cMCF7/XBP1

p< 0.001 for ANOVA,*p=0.029^p=0.019

^

*

XBP1 siRNA reduces BCL2

Rebecca Riggins et al., Mol Cancer Ther, 2005Bianca Gomez et al., FASEB J, 2007

Inhibition of both BCL2 and BCLW is better BECN1 (siRNA) and 3-MA each reverse antiestrogen resistance when combined with BCL2 (YC137) inhibition

apoptosisproliferation

Anatasha Crawford et al., PLoS ONE, 2010 Yanxia Ning et al., Mol Cancer Ther, 2010


Lombardi

BCL2 and Total-BH3 Predicts Level of Apoptosis

d[BAX ]

dt (k f 1 k f 2[BH3])[BAX] kb[BAXm]F

kb[BAXm :BCL2]

d[BAXm :BCL2]

dtkas[BAXm]F [BCL2]F kds[BAXm :BCL2]

kb[BAXm :BCL2]

d[BH3]Fdt

ks ksStress kd[BH3]F kasBH 3[BH3]F [BCL2]F

kdsBH 3[BH3 :BCL2]

d[BH 3 :BCL2]

dtkasBH 3[BH 3]F [BCL2]F kdsBH 3[BH 3 :BCL2]

kd[BH 3 :BCL2]

Mathematical Modeling: task = explore role of BCL2 family in apoptosis

Bill Bauman, Tongli Zhang in preparation

Model predicts %apoptosis and provides an approximate measureof responsiveness based on the concentrations of BCL2 and

the total of all BH3 members of the BCL2 family

PCD = programmed cell death/apoptosis

17 nonlinear ordinary differential equationsand 44 parameters for the various molecular species


Lombardi

Coordinated Functions: BCL2 Family and Cell Fate

Apoptosis(cell death)

Autophagy(cell survival)

altered cell metabolism?


Lombardi

Autophagy (self-eating)

Normal process through which aged or damaged subcellular organellesare degraded and their components recycled into intermediate

cellular metabolism

BECN1 (siRNA) and 3-MA (inhibit autophagy) reverse antiestrogen

resistance when combined with BCL2 (YC137) inhibition

Anatasha Crawford et al., PLoS ONE, 2010


Lombardi

Vehicle ICI 182, 780

Monodansylcadaverine-labeled Vesicles

XBP1(s) Induces Pro-Survival Autophagy

LC3-GFP expression

MCF7/XBP1 Vehicle MCF7/XBP1 1uM FAS

MCF7/EV 1uM Fas MCF7/EV Vehicle

Ayesha Shajahan et al., submitted


Lombardi

· Metabolome: collection of metabolites (~2500 identified in humans) e.g., within a cell

– reflects the physiological state of a cell

· Intermediates and products of metabolism (<1 kDa in size)e.g., amino acids, antioxidants, nucleotides, sugars, etc.

· Metabolites separated by mass and charge using UPLC-MS (Ultra Performance Liquid Chromatography-Mass Spectrometry)

· Data processed using Random Forest algorithm to identify most robust discriminant metabolites

Coordinated Functions: Metabolism

How does a cell coordinate its resources to allow execution of a cell fate decision?


Lombardi

High Confidence Interaction Network

METABOLITE

GENE/PROTEIN

MET. – PROT./MET.

PROT. – PROT.

Mapping metabolome onto transcriptome (LCC1 vs. LCC9)

Insulin/IGF signaling

Cell survival signaling

Energy metabolism



Lombardi

Antiestrogens Reduce Intracellular ATP

ATP levels drop with treatment in sensitive cells

Resistant cells have lower basal ATP levels that are refractory to endocrine treatment

LCC1 LCC9

AT

P le

vels

rel

ativ

e to

LC

C1

Veh

icle

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Vehicle E2 TAM FAS PAC

ATP

Vehicle=ethanol and no E2E2=17β-estradiolTAM=TamoxifenFAS=Fulvestrant/FaslodexPAC=Paclitaxel



Lombardi

Complete medium Glutamine (no glucose) medium

MYC, Glutamine, and UPR Enable LCC9 Survival

Ayesha Shajahan (GU) et al., submitted

UPR Activation


Lombardi

Cellular Sensing of Nutrient/Energy Deprivation

GRP78 and AMPK may be energy sensors and autophagy switches

BCL2 BCL2:BECN1

XBP1

XBP1 BCL2:BECN1 may confer degenerancy on autophagy induction

Katherine Cook et al., Cancer Res, 2012


Lombardi

A Mechanistic Topology of Endocrine Resistance

Clarke et al., Cancer Res, 2012Katherine Cook et al., Cancer Res, 2012

Cellular metabolism may be an essential determinant of cell fateor Glutamine (poor vascularization; loss of growth factor stimulation, etc.)

GRP78 = HSPA5 = BiP

BCL2, et al.

BECN1

Apoptosis

UPR

Autophagy


Lombardi

MetabolicAdaptations

System Coordination: Network Modeling

John Tyson et al., Nature Rev Cancer, 2011


Lombardi

Summary

● Systems biology approaches provide one way to explore phenotypes and to integrate cellular and molecular features to understand mechanism(s)

● Cells appear to experience EnR stress and can use GRP78 to activate the UPR, which then integrates signaling to determine cell fate

— inhibits apoptosis (e.g., antiapoptotic BCL2 family members)— induces autophagy (e.g., BECN1, antiapoptotic BCL2 family members, AMPK,

mTOR) — initiates/coordinates changes in metabolism required to execute the cell fate

decision

● Antiestrogens modify cellular energy metabolism leading to changes in glutamate/glutamine/glucose uptake and intracellular AMP levels

— autophagy also provides intermediate metabolites to fuel the cell fate decision

● ER acts as a bistable switching mechanism to affect phenotype, making

intermittent therapy a more effective strategy

● Some early adaptations to treatment are retained in resistant cells

● Resistance may not require many new nodes but does change the nature/usage of existing edges among nodes

(it’s mostly the same network of nodes, its just wired differently)


Lombardi

Acknowledgments

J. Michael Dixon University of Edinburgh, Breast Unit William R. Miller University of Edinburgh, Breast UnitLorna Renshaw University of Edinburgh, Breast UnitAndrew Simms University of Edinburgh, Breast UnitAlexey Larionov University of Edinburgh, Breast Unit

Bill Baumann Engineering & Computer ScienceChun Chen Engineering & Computer ScienceLi Chen Engineering & Computer ScienceIman Tavasolly Biological Sciences & Virginia Bioinformatics InstituteJohn Tyson Biological Sciences & Virginia Bioinformatics InstituteAnael Verdugo Biological Sciences & Virginia Bioinformatics InstituteYue Wang Engineering & Computer ScienceJianhua Xuan Engineering & Computer ScienceBai Zhang Engineering & Computer Science

Erica GolemisRochelle NastoIlya Serebriiskii

Harini Aiyer Amrita Cheema Sandra JablonskiYounsook Cho Katherine Cook Yongwei ZhangAhreej Eltayeb Caroline Facey Lou WeinerLeena Hilakivi-Clarke Rong Hu Subha MadhavanMike Johnson Lu Jin Yuriy GusevHabtom Ressom Rebecca B. Riggins Robinder GaubaJessica Schwartz Ayesha Shajahan Minetta Liu (now at Mayo)Anni Wärri Alan Zwart

U54-CA149147 ICBP Center for Cancer Systems Biology

29XS194 NCI In Silico Research Center of Excellence

R01-CA131465; R01-CA149653

KG090245BC073977BC122874

The patients who contributed to the clinical studies

Modeling Endocrine Resistance in Breast Cancer Robert Clarke, Ph.D., D.Sc. Professor of Oncology Director, Center for Cancer Systems Biology Dean for Research.

Documents

gfr gfr activation

gfr dependence gfr

gfr rate of gfr

gfr basal inhibition

gfr epi activation

erp gfr activation

e2er gfr inhibition

gfrover gfr activation