Upload
madison-wade
View
225
Download
2
Tags:
Embed Size (px)
Citation preview
CELL MODEL CALIBRATION
Using NMR based metabonomics
Kranthi VaralaAdvisor : Peter OrtolevaCapstone Project Spetember 2004
Kranthi Varala - Capstone Project
Cell models and simulators
Cell models study cell behavior and cell response
Powerful predictive tools
Simulators have to be able to predict behavior
accurately
Stochastic, Flux balance and Kinetic simulators exist
Kranthi Varala - Capstone Project
Background – Karyote
Karyote is a compartmentalized, kinetic cell
simulator (http://ruby.chem.indiana.edu)
Kinetic model superior to stochastic models
(Gillespie solutions) and Flux balance analysis
Harder to build and calibrate model
Kranthi Varala - Capstone Project
Motivation
Utilization of NMR data
Adapt our information theory approach to use
established experimental measurements
Utilization of multiplex data
Concurrent usage of different kinds of data
Kranthi Varala - Capstone Project
Nuclear Magnetic Resonance (NMR)
Chemicals (metabolites) with 13C can be detected
Position of the peaks is always constant and unique for
a given molecule
Position marked in ppm (ratio from original signal)
Inte
nsity
13C Spectrum for Toluene (http://www.cis.rit.edu/htbooks/nmr/inside.htm)
Kranthi Varala - Capstone Project
NMR based metabonomics
Intensity of peak is measure of its concentration in sample
Recent advances in NMR enhanced amplitude sensitivity
Reproducibility in the range of +/- 0.2-1.0% is reported in
the NMR community
Single cell isolation techniques help separation of a single
cell which can then be ruptured and its contents sampled
Kranthi Varala - Capstone Project
Spectrum complexity increases rapidly Dense spectra often have overlapping peaks Inversion of spectrum to metabolite
concentrations difficult
13C spectrum of Mountain DewImage: www.acts.org/roland/mt.dew
1H spectrum of one protein
Spectrum Complexity
Kranthi Varala - Capstone Project
Current approaches to NMR based Metabonomics Many papers published recently deal with the inversion problem. Deconstructing the
spectral intensities into concentrations.
Pre-processing spectrum
Normalization
Remove water, TMSP etc. peaks
Log scaling
Statistical analysis
Multivariate Analysis
Molecular Factor analysis
Most solutions computationally intensive
Kranthi Varala - Capstone Project
Simplification of spectral complexity
1H spectra are too dense to process. 13C spectra sparser but still
overwhelming
13C spectra have a wider spectral range(~200ppm) compared to 1H
(~15ppm)
Our solution is to grow cells in 13C enriched media to enhance 13C
spectra which are inherently sparse
Spectra from these cells will show peaks only for those metabolites that
are synthesized through metabolism using 13C medium components
Kranthi Varala - Capstone Project
Avoiding inversion
Faster processing, less computation
Generate synthetic NMR from metabolite concentrations
Spectral database for common metabolites
Predicted concentrations from Karyote translated to
spectrum
Kranthi Varala - Capstone Project
Synthetic NMR - Our approach
Conversion factor is provided by addition of a reference
compound
Known concentration of reference compound carefully added
to sample prior to data acquisition
Concentration of metabolite peaks computed as ratio against
the reference peak
Kranthi Varala - Capstone Project
Parameters in Karyote
List of parameters in Karyote
Initial concentration of metabolite
Rate of reaction
Equilibrium constant of reaction
Rate of transport across membrane
Kranthi Varala - Capstone Project
Calibration
Measure of intracellular metabolite levels gives
valuable information to calibrate a cell reaction-
transport model
Time series data ideal, discrete data can also be used
Information theory calibrates model by adjusting
parameters iteratively
Kranthi Varala - Capstone Project
Information theory (IT)
Probability based formulation to
calibrate cell model (Sayyed-
Ahmad et al. 2003)
Error minimization techniques to
calibrate kinetic parameters
Uncertainty of the system is
limited only by available data
In principle different data types
can be used in error computation
Kranthi Varala - Capstone Project
IT workflow
Kranthi Varala - Capstone Project
Making IT modular
IT built to use direct metabolite concentration data
Make each data type a module
Code optimized towards this end. IT can now accept any
kind of data if its parsing and error computation modules
are provided
NMR developed as a module for the core IT program
Kranthi Varala - Capstone Project
IT using NMR data
Kranthi Varala - Capstone Project
NMR error computation
Comparison of 2 spectra as line data
Inherently simplifies the spectrum by ignoring lines that need not be
compared
Allows computation without complete knowledge of spectra for all
species in the cell
Error computed as difference between synthetic and experimental
NMR
Kranthi Varala - Capstone Project
Error surface
Kranthi Varala - Capstone Project
Error surface-oscillatory model
Kranthi Varala - Capstone Project
Dual data schemes and cross-cell analysis
One cell model can be used to understand another less understood but
related cell
Spectral data obtained from both cells and processed to discover the
underlying functional differences between the two networks
Algorithm starts with the defined cell model and adjusts parameters on
subsequent iterations to match the spectrum for the new cell type
Typical example is comparing a normal to a mutated cell
Comparison between two organisms is also plausible
Kranthi Varala - Capstone Project
ResultsCell Model Parameter Initial guess Optimized value Correct Value Error %
4 metabolites, 2 reaction & transport model
Equillibrium Constant
1e-4 3e-4 3e-4 - 9e-3 0
4 metabolites, 2 reaction & transport model
Rate of reaction 0.0001 0.01 0.01 0
7 metabolites, oscillatory model
Rate of reaction 1e-4 2.5e-4 1e-5 96
Trypanosoma Equilibrium constant
1 110 126.41 13
Trypanosoma Rate of reaction 1e-9 1.28e-10 1.4e -10 8.57
Kranthi Varala - Capstone Project
ReferencesSayyed-Ahmad A, Tuncay K, Ortoleva P. American Chemical Society, Jun 30 2003
Sterin M, Cohen S, Mardor Y, Berman E, Ringel I. Cancer Research 61, Oct 15 2001
Eads C, Furnish C, Noda I, Juhlin K, Cooper A, Morrall S. Analytical Chemistry., 76(7)
Mar 9 2004
Lenz E.M, Bright J, Wilson I.D, Morgan S.R, Nash A.F.P Journal of Pharm. And
biomed. Anal. 33(5) Dec 5 2003
Reo N.V. Drug and chemical toxicology 25(4) 2002
Atlas of Carbon-13 NMR data Breitmaier E, Haas G, Voelter. W. Heyden & Son, 1979
13 C NMR spectroscopy Breitmaier E, Voelter W. Verlag Chemie 1974
The Aldrich library of 13 C and 1 H FT-NMR spectra. Pouchert C.J, Behnke J, Aldrich
chemical company Inc. 1993
Kranthi Varala - Capstone Project
Acknowledgements
Peter Ortoleva
Sun Kim
Abdallah Sayyed-Ahmad
Haixu Tang
John Tomaszewski