PROJECT OUTLINE

PROSTATE CANCER

M. Koletou1, 2, 3, M. Gabrani2, T. Guo3, Q. Zhong1, U. Wagner1, R. Aebersold3, P. Wild1, M. Rodríguez Martínez2

1 Institute of Pathology and Molecular Pathology, University Hospital Zurich, Switzerland 2 IBM Research Laboratory Zurich, Switzerland 3 Institute of Molecular Systems Biology, ETH Zurich, Switzerland

First and foremost we would like to acknowledge SystemsX.ch whose funding made this interdisciplinary

PhD project possible. We thank The Cancer Genome Atlas (TCGA) Network for granting us access to the

Prostate Adenocarcinoma datasets. We also want to thank the SystemsX.ch project PhosphoNet

Personalized-Precision Medicine (SystemsX.ch project no. 2012/191), for the genomic and proteomic data

that are made available to us. Finally, we would also like to show our gratitude towards the University

Hospital of Zurich, IBM Research Laboratory Zurich and ETH Zurich for their amicable collaboration and their

expertise that greatly assist this project.

epsilon xD - Ys, x.. , xD - Ymin F0

2ts

Acknowledgments

A COMPUTATIONAL FRAMEWORK FOR SYSTEMS PATHOLOGY OF PROSTATE CANCER

[1] de Morsier, F. et al., 2014. In K. Lai & A. Erdmann, eds. SPIE Advanced Lithography. International Society for Optics and Photonics, p. 905211.

The 2016 WHO Classification of Tumors of the

Urinary System and Male Genital Organs-Part B:

Prostate and Bladder Tumors.European urology.

• Grade group 1: Gleason score ≤ 6

• Grade group 2: Gleason score 3 + 4 = 7

• Grade group 3: Gleason score 4 + 3 = 7

• Grade group 4: Gleason score 8

• Grade group 5: Gleason scores 9–10

*Age-adjusted to the 2000 US standard population and adjusted for relays in reporting.†Includes the intrahepatic bile duct

Sourec: Surveilance, Epidemiology, and End Results (SEER) Program, National Cancer

Institute, 2014.

image source: American Cancer Society, Cancer Facts & Figures 2015

Incidence and Significance

• very high incidence

• but most cases are insignificant

• diagnostic screening is controversial

• overdiagnosis and overtreatment

Need for biomarker candidates

• stratify aggressive from insignificant PCa

• more accurate prognosis

− better than Gleason score (a histopathological

grading of prostate tissue obtained by biopsy)

NOVEL COMPUTATIONAL

FRAMEWORK

stratify prostate cancer into

insignificant and aggressive

identify genomic alteration

profiles that enable the

stratification

integrate multi-omics datasets

to strengthen the analysis

D G

D

P

G

(i) sparse

dictionary learning

(ii) network

visualization

(iii) mapping dictionary to

phenotypic traits and

genomic alterations

Molecular Dataset Dictionary Dictionary NetworkPhenotype - Genotype

association network

raw genomic

datasets

molecular fingerprints

of cancer

PD

DICTIONARY LEARNING WITH SPARSE CODING[1]

Pathway Example: HALLMARK PI3K-AKT-MTOR SIGNALING

pathway network from Data correlations

DATA: Copy Number Alteration (CNA) profilesfrom TCGA prostate tumor samples

Dictionary (D)

Sparse Coefficients (X)

pathway network from Dictionary sparse correlations

Data

Dictionary

Comparison of

Betweenness Centrality

of the pathway networks

the number of the shortest paths

from node to node

that pass through node

Betweenness Centrality Betweenness Centrality

Co

rre

lati

on

Co

rre

lati

on

Dictionary Mapping

Phenotype – Genotype association with Dictionary Learning

patient to gene

specific mapping

allows us to explore a

more personalized

molecular fingerprint of

prostate cancer

all pathway genes in the CNA dataset

selected genes for all patients with dictionary mapping

selected genes for each patient with dictionary mapping