Transcript Presentation

Software for
Modeling and Simulation
of Biochemical Networks
http://www.copasi.org
Stefan Hoops
Virginia Bioinformatics Institute
Overview
• Motivation
• COPASI
• Acknowledgement
• Other Software
Biochemical Networks
Simple Network:
Data:
A
B
C+D
E
B+2F
G +H
Modeling Paradigm
• Top Down
• Phenomenological modeling
approach describing
experimental data.
• Bottom Up
• Small well understood
models, e.g. enzymatic
reactions are used to comprise
larger models.
What we need?
• Easy to use analysis tools
• Interaction between tools
• Simulation with trusted results
• Parameter estimation capabilities
• Model comparison.
COPASI Features
• Time Course
• Steady State
• Structural Analysis
• Metabolic Control Analysis
• Lyapunov Exponent Calculation
• Parameter Scan
• Optimization
• Parameter Fitting
ODE Based Time Course Simulation
.
A.
B.
C.
D.
E.
F.
G.
H
-1
1
0
= 0
0
0
0
0
0
0
-1
-1
1
0
0
0
In general:
0
-1
0
0
0
-2
1
1
.
v 1 (A, B, H)
v 2 (B, C, D, E)
v 3 (B, E, F, G, H)
x = N v with: x =
x1
x2
.
.
.
xn
v1
v2
v=
.
.
.
vm
Stochastic Time Course Simulation
Initialize system
Calculate: Reaction probabilities
Generate random numbers to determine:
• time of next reaction
• which reaction happens
Update the system the system
Example
Optimization
Optimization attempts to maximize or
minimize an objective function.
• Note, that the maximum of a function f is
equivalent to the minimum –f
• Given a real-valued scalar function f(x,k)
of n parameters k=(k1, ..., kn) find a
minimum of f(x,k) such that:
• gi(x) ≥ 0 with i=1,..., m (inequality constraints)
• hj(x) = 0 with j=1,..., m’ (equality constraints)
Numerical Optimization Cycle
Optimization Methods
• Gradient based
• Steepest Descent
• Levenberg Marquard
• Direct
• Deterministic
• Hooke & Jeeves
• Random
• Genetic Algorithm
• Evolutionary Programming
• Random Search
• Nelder Mead
• SRES
• Simulated Annealing
Parameter Estimation / Fitting
This is a case of
optimization with a special
objective function:
The simulation results
shall match the
experimental results
closely.
Parameter Estimation Result
Command Line Interface
• Suitable for long computational task like
Optimization or Parameter Estimation
• Background progress for Web-applications or
Web-services
• Basic usage:
• Create a model with the COPASI GUI
• Specify computational task in the GUI
• Save File “model.cps”
• CopasiSE “model.cps”
Available Platforms
• Linux
• All WIN32 OS starting Windows 98 (Intel)
• Mac OS X (PowerPC and Intel)
• SunOS starting with Solaris 8 (sparc)
• Achieved through
• QT (Toolkit and libraries for GUI development)
• LAPACK / BLAS (matrix and vector routines)
• ODEPACK (ODE solver)
• EXPAT (XML library)
• LIBSBML (SBML library)
Availability
• Current Release (June 2006)
COPASI Version 4.0 Build 18
• COPASI is publicly available since
October 2004 (Build 9)
Community Integration
• SBML import and export
• Berkeley Madonna export
• C source code generation
Acknowledgements
Mendes group @ VBI
Pedro Mendes:
Sameer Tupe:
Anurag Srivastava:
Christine Lee:
Gaurav Singh:
Mrinmyee Kulkarni:
Liang Xu:
Mudita Singhal:
Rohan Luktuke:
Ankur Gupta:
Wei Sun:
Yonqun (Oliver) He:
Aejaaz Kamal:
Principal Investigator, occasional programmer, tester, and webmaster
Programmer (Fall 2004 - Fall 2005)
Programmer (Fall 2004 - Summer 2005)
Programmer (Fall 2003 - Spring 2005)
Programmer (Fall 2003 - Spring 2004)
Programmer (Spring 2002 - Fall 2003)
Programmer (Spring 2003 - Fall 2003)
Programmer (Spring 2002 - Summer 2003)
Programmer (Summer 2002 - Fall 2002)
Programmer (Spring 2002 )
Programmer (Fall 2001 - Summer 2002)
Programmer (Fall 2001 - Spring 2002)
Programmer (Spring 2001 - Summer 2001)
Kummer group @ EML Research
Ursula Kummer:
Principal Investigator, tester
Sven Sahle:
Software architect, project manager, programmer
Ralph Gauges:
Software engineering, programmer, documentation
Juergen Pahle:
Programmer
Natalia Simus:
Programmer
Jürgen Zobeley:
Tester
Ursula Rost:
Programmer
Katja Wegner:
Tester, programmer, documentation
Ralph Voigt:
Documentation
Sarah Lilienthal:
Programmer (July - August 2005)
Wenjun Hu:
Programmer (August 2003 - October 2003)
Carel van Gend:
Programmer (October 2000 - May 2002)
DOME
• DOME is a database and analysis system for functional genomics
projects.
• It can be used to store and analyze transcriptomics, proteomics, and
metabolomics data.
• The analysis that can be performed with DOME allow for an
integrated view of the data generated using different technologies.
• We have implemented the system on three functional genomics
projects on Medicago truncatula, Vitis vinifera and Saccharomyces
cerevisiae and thus have attempted to make the system general
enough to be used by various labs for their functional genomics
needs.
Overview of DOME
Microarray
Data storage
and processing
Sampling_replicate
ma_normalized
2D-PAGE
protein_normalized
GC/MS; LC/MS; CE/MS
metabolite_normalized
Experiment
metadata
Sampling_point
sp_summary
gene
protein
B-Net
compound
Data analysis
Statistical Analysis
- Unsupervised (PCA, clustering)
- Supervised (Discriminant analysis, GA-MDA, and others)
event
Visualization
- Biochemical Maps
(using BROME)
Multivariate Data Analysis for
Genomics and Systems Biology
• Current analyses provided:
• correlation analysis
• partial correlation analysis
• principal component analysis (PCA), including biplot
display
• linear multiple discriminant analysis (MDA),
• linear multiple discriminant analysis with genetic
algorithm variable selection (GA-DFA) - 2 different
algorithms.
• non-negative matrix factorization (NMF)