Transcript Multiple perturbation analysis of cancer pathways

How about a Master’s project in the
area of cancer systems biology?
• We offer projects in the following areas:
• Reverse engineering of cancer pathways
• Control of the cancer cell phenotype using drug
combinations
• Predicting phenotype responses to cellular perturbation
from array data
• Contact for more info:
• [email protected] (please attach a CV)
• Previous project workers:
•
•
•
•
Frank Eriksson, Mat Stat, Chalmers
Darima Lamazhapova, Cambridge University
Erik Larsson, Wallenberg lab, GU
Tanya Lobovkina, Chemistry, Chalmers
Multiple Perturbation Analysis
of Cancer Pathways
Sven Nelander
Computational Biology Center / Chris Sander group
Memorial Sloan-Kettering Cancer Center
New York
Outline
• Perturbing cancer cells - key questions
• Building models for combinatorial perturbation of
breast cancer pathways
• Whole-genome RNAi screening to extend the
TGF-beta pathway
• Work in cancer genomics and related areas
Cancer
• A broad class of diseases exhibiting uncontrolled
growth, tissue invasion and metastasis.
• Gradual progression towards a more malignant
phenotype
• Acquisition of mutations that affect a specific set of
processes
•
•
•
•
•
Growth factor signaling
Apoptosis
DNA repair
Cell cycle regulation
Differentiation
Selective targeting of cancer pathways by small
compounds
1. Force differentiation
2. Inhibit anti-apoptotic signals
3. Inhibit growth-stimulating signals
Tumors contain multiple genetic abnormalities
•
Copy number alterations (MSKCC study).
Gain of DNA
200 Patients
Loss of DNA
genomic position (3000 megabases)
• Sequence alterations. 90 mutated genes per tumor in breast
and colon cancer (Sjöblom et al, Science 2006)
• Promoter hypermethylation.
A role for systems biology?
• Key types of question:
•
•
•
•
How will a melanoma cell line with mutation X respond to drug Y?
Will drug X synergize with drug Y?
Which regulatory interactions are implicated by the observed responses?
What’s the mechanism of action of drug X?
• Pathway maps are ambiguous, incomplete and have unclear
predictive value.
• Expert intuition is likely to fail in complicated cases.
• To facilitate prediction and inference, mathematical models can be
employed.
Implementing a systems biology cycle for combined perturbations
Desirable properties of a model for our
purposes
Representational capability:
•
Pathway-like and biochemically plausible
•
Quantitative or semi-quantitative predictions
•
Nonlinear interaction effects (epistasis and synergism) are possible
Experimental implications:
•
Both temporal and steady state perturbation responses
•
Incomplete readout possible
Algorithms:
•
Reverse engineering is computationally tractable
Simple dynamical models
dxi
 (W ij x j )   i x i  Pi
dt
j

Similar models used for
Analysis of microarray time series (D’Haeseler 2000, Xiong 2004)
Network inference from perturbed microarray profiles (Yeung 2002, Tegner 2003)
Inference of mechanism of action (diBernardo et al 2006)
dxi
  i f (W ij x j  Pi )   i x i
dt
j

Similar models used for
Analysis of microarray time series (D’Haeseler, 2000),
Modeling of lambda phage gene regulation (Vohradsky, 2001),
Robustness analysis of the yeast cell cycle (Li et al 2004).
Discussed as a model for signaling in (Bhalla 2003).
DNA switch network - synthetic biology (Kim et al 2004 and 2006)
Prediction of perturbation responses
COMPOUNDS
PHENOTYPE
Prediction of perturbation responses
Experimental data from Kaufman et al, PLoS Comp Biol, 2006
Parameter fitting / system identification
• For all experiments minimize
E  E SSQ  E STRUCT

Sum of squares
error
Solmaz Shahalizadeh, Master’s thesis
Structural
complexity
Algorithms used to minimize E
• Recurrent backpropagation (Pineda, 1988)
• Backpropagation through time (Pearlmutter)
• Gennemark and Wedelin, 2007
Inference from steady state perturbation responses,
hypothetical experiment with 40 dual perturbations
and 10 readouts
Inference from perturbation
responses, experimental data
Data from Janes et al, Science 2005
Experimental pilot studies (ongoing)
• Two breast epithelial cell lines
• MCF7 - cancer
• MCF10A - transformed noncancer
• Initial focus on mitogenic pathways and low
molecular weight compound perturbation
• Database of 2200 compound-gene links
• Experiment 1: predict triplet perturbations from dual perturbations
• Experiment 2: crosstalk detection and explanation
Efficient proteomics technique will make
large perturbation studies possible.
SILAC technology (Jens Andersen group, Odense)
Reverse phase protein array (Weiqing Wang)
Dilution of Lysate
0
1/2
1/4
1/8
Duplicates
0
1/2
1/4
1/8
1/16 1/32 1/64 1/128
1/16 1/32 1/64 1/128
Duplicates
1) One grid for one sample
2) One antibody blot for one slide
3) Relative quantification, positive
controls on each slide
4) Quantitative peptide and
phosphopeptide controls
RNAi screening for TGF-beta pathway components
(Niki Schultz)
21000 siRNA duplexes were scored for their effect on TGFbeta signaling
Work in genomics
• Sarcoma genome project
• Collaboration with MSKCC surgery dept and Broad Inst.
• 140 sarcoma patients
• Large-scale genomic characterization:
– Transcriptional arrays
– Copy number arrays
– Exon sequencing
• Aberrant processes? Therapy targets?
• DNA copy number alteration in nonmalignant lesions
• Collaboration with Columbia pathology dept.
Summary
• Methodology to analyze combinatorial perturbation
experiments using differential equation models.
• Preliminary data suggest applicability to real experimental data
• No assumptions of linearity or complete observation
• The methodology generalizes genetic epistasis analysis in that
it handles higher order perturbations and feedback loops.
• We are proceeding to a study of combinatorial drug
effects on the phenotype of breast cancer cells.
Future perspectives
• Using perturbation to pinpoint mutations and
regulatory differences between tumors
• Cancer genomics data as an endogenous
perturbation experiment
• Phenotype control in non-malignant disease
conditions
Acknowledgements
• Weiqing Wang, Nikolaus Schultz,
Christine
Pratilas, Barry Taylor, Dina Marenstein, Sam
Singer, Joan Massague, Neal Rosen, Chris
Sander
• Solmaz Shahalizadeh, Peter Gennemark,
Frank Eriksson, Darima Lamazhapova
• Søren Schandorff, Jens Andersen
• Björn Nilsson