Key Points
-
Mathematical modelling is an essential tool in systems biology. To ensure the accuracy of mathematical models, model parameters must be estimated using experimental data, a process called regression. Also, pre-regression and post-regression diagnostics must be employed to evaluate the model goodness-of-fit and the reliability of the estimated parameter values.
-
Maximum likelihood estimation and least-squares fitting are the most common regression schemes, yielding parameter values and their variance–covariance matrix. They work under the assumption that the estimated parameters have a normal distribution. When this assumption is not valid, Bayesian inference can be used, yielding the full parameter distribution.
-
Prior to regression, the structural identifiability of models must be assessed to determine whether model parameters can be uniquely determined and what data are required to achieve that.
-
Post-regression diagnostics include testing a model's goodness-of-fit, determining which model among competing ones fits the data best, evaluating parameter determinability and evaluating parameter significance.
-
Parameters in probabilistic models must be inferred by either indirect inference or by Bayesian methods. In indirect inference, model parameters are estimated by minimizing the differences between intermediate statistics that characterize simulated and experimental data.
Abstract
Mathematical models are an essential tool in systems biology, linking the behaviour of a system to the interactions between its components. Parameters in empirical mathematical models must be determined using experimental data, a process called regression. Because experimental data are noisy and incomplete, diagnostics that test the structural identifiability and validity of models and the significance and determinability of their parameters are needed to ensure that the proposed models are supported by the available data.
This is a preview of subscription content, access via your institution
Access options
Subscribe to this journal
Receive 12 print issues and online access
$189.00 per year
only $15.75 per issue
Buy this article
- Purchase on Springer Link
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
References
Arkin, A. P. Synthetic cell biology. Curr. Opin. Biotechnol. 12, 638–644 (2001).
Kitano, H. Systems biology: a brief overview. Science 295, 1662–1664 (2002).
Sachs, K., Perez, O., Pe'er, D., Lauffenburger, D. A. & Nolan, G. P. Causal protein-signaling networks derived from multiparameter single-cell data. Science 308, 523–529 (2005).
Woolf, P. J., Prudhomme, W., Daheron, L., Daley, G. Q. & Lauffenburger, D. A. Bayesian analysis of signaling networks governing embryonic stem cell fate decisions. Bioinformatics 21, 741–753 (2005).
Bulashevska, S. & Eils, R. Inferring genetic regulatory logic from expression data. Bioinformatics 21, 2706–2713 (2005).
Segal, E., Friedman, N., Kaminski, N., Regev, A. & Koller, D. From signatures to models: understanding cancer using microarrays. Nature Genet. 37, S38–S45 (2005).
Janes, K. A. et al. Systems model of signaling identifies a molecular basis set for cytokine-induced apoptosis. Science 310, 1646–1653 (2005).
Janes, K. A. et al. Cue-signal-response analysis of TNF-induced apoptosis by partial least squares regression of dynamic multivariate data. J. Comp. Biol. 11, 544–561 (2004).
Heard, N. A., Holmes, C. C., Stephens, D. A., Hand, D. J. & Dimopoulos, G. Bayesian coclustering of Anopheles gene expression time series: study of immune defense response to multiple experimental challenges. Proc. Natl Acad. Sci. USA 102, 16939–16944 (2005).
Sprague, B. L. et al. Mechanisms of microtubule-based kinetochore positioning in the yeast metaphase spindle. Biophys. J. 84, 3529–3546 (2003).
Bentele, M. et al. Mathematical modeling reveals threshold mechanism in CD95-induced apoptosis. J. Cell Biol. 166, 839–851 (2004).
Gardner, M. K. et al. Tension-dependent regulation of microtubule dynamics at kinetochores can explain metaphase congression in yeast. Mol. Biol. Cell 16, 3764–3775 (2005).
Rodriguez-Fernandez, M., Mendes, P. & Banga, J. R. A hybrid approach for efficient and robust parameter estimation in biochemical pathways. Biosystems 83, 248–265 (2006).
Mendes, P. & Kell, D. B. Non-linear optimization of biochemical pathways: applications to metabolic engineering and parameter estimation. Bioinformatics 14, 869–883 (1998).
Schoeberl, B., Eichler-Jonsson, C., Gilles, E. D. & Muller, G. Computational modeling of the dynamics of the MAP kinase cascade activated by surface and internalized EGF receptors. Nature Biotechnol. 20, 370–375 (2002). Ordinary differential equation-based model of the epidermal-growth-factor-signalling network with parameters that were estimated using sensitivity analysis and least-squares regression of concentration time-courses.
Bellman, R. & Astrom, K. J. On structural identifiability. Math. Biosci. 7, 329–339 (1970).
Yao, K. Z., Shaw, B. M., Kou, B., McAuley, K. B. & Bacon, D. W. Modeling ethylene/butene copolymerization with multi-site catalysts: parameter estimability and experimental design. Polymer Reaction Eng. 11, 563–588 (2003).
Gadkar, K. G., Gunawan, R. & Doyle, F. J. III Iterative approach to model identification of biological networks. BMC Bioinformatics 6, 155 (2005). Presents details of structural identifiability analysis and its application to parameter estimation and optimal experimental design.
Hastie, T., Tibshirani, R. & Friedman, J. The Elements of Statistical Learning — Data Mining, Inference and Prediction (Springer, New York, 2001).
Papoulis, A. in Probability, Random Variables, and Stochastic Processes (ed. Editions, M.-H. I.) (McGraw-Hill, New York, 1991).
Press, W. H., Flannery, B. P., Teukolsky, S. A. & Vetterling, W. T. Numerical Recipes in C: The Art of Scientific Computing (Cambridge Univ. Press, New York, 1992).
Golub, G. H. & Van Loan, C. F. An analysis of the total least squares problem. SIAM J. Numer. Anal. 17, 883–893 (1980).
Danuser, G. & Strickler, M. Parametric model fitting: from inlier characterization to outlier detection. IEEE Trans. Patt. Anal. Mach. Intell. 20, 263–280 (1998).
Rousseeuw, P. J. Least median of squares regression. J. Am. Stat. Ass. 79, 871–880 (1984).
Koch, K. -R. Parameter Estimation and Hypothesis Testing in Linear Models (Springer, Berlin, 1988).
Pardalos, P. M. & Romeijn, H. E. Handbook of Global Optimization Volume 2 (Kluwer Academic, Dordrecht, 2002).
Horst, R. & Pardalos, P. M. Handbook of Global Optimization (Kluwer Academic, Dodrecht, 1995).
Moles, C. G., Mendes, P. & Banga, J. R. Parameter estimation in biochemical pathways: a comparison of global optimization methods. Genome Res. 13, 2467–2474 (2003).
Kleinbaum, D. G., Kupper, L. L., Muller, K. E. & Nizam, A. Applied Regression Analysis and Multivariable Methods (Duxbury, 1997).
Seber, G. A. & Wild, C. J. Nonlinear Regression (Wiley-Interscience, Hoboken, 2004). References 29 and 30 are comprehensive textbooks on linear and nonlinear regression and important related diagnostics.
Efron, B. Nonparametric estimates of standard error: the jackknife, the bootstrap and other methods. Biometrika 68, 589–599 (1981).
Potvin, C. & Roff, D. A. Distribution-free and robust statistical methods: viable alternatives to parametric statistics. Ecology 74, 1617–1628 (1993).
Coleman, M. C. & Block, D. E. Bayesian parameter estimation with informative priors for nonlinear systems. AIChE J. 52, 651–667 (2005).
Barenco, M. et al. Ranked prediction of p53 targets using hidden variable dynamic modeling. Genome Biol. 7, R25 (2006).
Chen, M., Shao, Q. & Ibrahim, J. G. Monte Carlo Methods in Bayesian Computation (Springer, New York, 2000). Presents many computational techniques for carrying out Bayesian inference.
Schwarz, G. Estimating dimension of a model. Ann. Stat. 6, 461–464 (1978).
Gruen, A. W. Data-processing methods for amateur photographs. Photogramm. Rec. 11, 567–579 (1985).
Golub, G. H. & Van Loan, C. F. Matrix Computations (Johns Hopkins Univ. Press, Baltimore, 1983).
Pedraza, J. M. & van Oudenaarden, A. Noise propagation in gene networks. Science 307, 1965–1969 (2005).
Rosenfeld, N., Young, J. W., Alon, U., Swain, P. S. & Elowitz, M. B. Gene regulation at the single-cell level. Science 307, 1962–1965 (2005).
Bennett, M. R. & Kearns, J. L. Statistics of transmitter release at nerve terminals. Prog. Neurobiol. 60, 545–606 (2000).
Redman, S. Quantal analysis of synaptic potentials in neurons of the central nervous system. Physiol. Rev. 70, 165–198 (1990).
Morton-Firth, C. J. & Bray, D. Predicting temporal fluctuations in an intracellular signalling pathway. J. Theor. Biol. 192, 117–128 (1998).
Spudich, J. L. & Koshland, D. E. Non-genetic individuality — chance in single cell. Nature 262, 467–471 (1976).
Mitchison, T. & Kirschner, M. Dynamic instability of microtubule growth. Nature 312, 237–242 (1984).
Smith, A. A. Jr Estimating nonlinear time-series models using simulated vector autoregression. J. Appl. Econometrics 8, S63–S84 (1993). Introduces methods of indirect inference for the estimation of parameters in probabilistic models.
Gourieroux, C., Monfort, A. & Renault, E. Indirect inference. J. Appl. Econometrics 8, S85–S118 (1993).
Jiang, W. & Turnbull, B. The indirect method: inference based on intermediate statistics — a synthesis and examples. Stat. Sci. 19, 239–263 (2004).
Gallant, A. R. & Tauchen, G. Which moments to match? Econometric Theory 12, 657–681 (1996).
Golightly, A. & Wilkinson, D. J. Bayesian sequential inference for stochastic kinetic biochemical network models. J. Comp. Biol. 13, 838–851 (2006).
O'Neill, P. D. & Roberts, G. O. Bayesian inference for partially observed stochastic epidemics. J. Royal Stat. Soc. A 162, 121–129 (1999).
Gibson, G. J., Kleczkowski, A. & Gilligan, C. A. Bayesian analysis of botanical epidemics using stochastic compartmental models. Proc. Natl Acad. Sci. USA 101, 12120–12124 (2004).
Smith, A. F. M. & Roberts, G. O. Bayesian computation via the Gibbs sampler and related Markov chain Monte Carlo Methods. J. Royal Stat. Soc. B 55, 3–23 (1993).
Wilkinson, D. J. Stochastic Modelling for Systems Biology (CRC Press, Boca Raton, 2006). Discusses issues that are related to probabilistic modelling and the estimation of parameters in stochastic models using Bayesian inference.
Walker, R. A. et al. Dynamic instability of individual microtubules analyzed by video light-microscopy — rate constants and transition frequencies. J. Cell Biol. 107, 1437–1448 (1988).
Shaw, S. L., Yeh, E., Maddox, P., Salmon, E. D. & Bloom, K. Astral microtubule dynamics in yeast: a microtubule-based searching mechanism for spindle orientation and nuclear migration into the bud. J. Cell Biol. 139, 985–994 (1997).
Odde, D. J., Cassimeris, L. & Buettner, H. M. Kinetics of microtubule catastrophe assessed by probabilistic analysis. Biophys. J. 69, 796–802 (1995).
Gildersleeve, R. F., Cross, A. R., Cullen, K. E., Fagen, A. P. & Williams, R. C. Microtubules grow and shorten at intrinsically variable rates. J. Biol. Chem. 267, 7995–8006 (1992).
Dorn, J. F. et al. Interphase kinetochore microtubule dynamics in yeast analyzed by high-resolution microscopy. Biophys. J. 89, 2835–2854 (2005).
Jaqaman, K. et al. Comparative autoregressive moving average analysis of kinetochore microtubule dynamics in yeast. Biophys. J. 91, 2312–2325 (2006).
Acknowledgements
This work was supported in part by a National Institutes of Health grant. K.J. is a Paul Sigler/Agouron fellow of the Helen Hay Whitney Foundation.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Supplementary information
Related links
Glossary
- Structural identifiability
-
A model is structurally identifiable if its parameters can be uniquely estimated by fitting the model to experimental data. Structural identifiability is related to the sensitivity of process output to parameter variations.
- Variance
-
A measure of the dispersion of a variable around its average. Its square root is the standard deviation.
- Covariance
-
A measure of how two variables vary relative to each other.
- Significance
-
A parameter is statistically significant if, given the uncertainty in its estimate due to noise in the input data, the probability that the parameter magnitude is different from zero not just by chance exceeds the confidence required by the investigator.
- Determinability
-
A measure for the capability to infer the value of a model parameter from the available input data, independent of the values of other parameters.
- Regression instability
-
A measure for the variation of regression results in the presence of data noise. A regression is unstable if the estimates of model parameters significantly differ when one additional data point is added to the set of input data.
- Linear independence
-
A set of parameters is linearly independent if none of its parameters can be written as a linear combination of the other parameters.
- Normal distribution
-
A bell-shaped distribution that is fully characterized by its mean μ and variance σ2. It is usually written as N(μ, σ2).
- Residual
-
The difference between an observation and the corresponding model prediction.
- Robust
-
An estimation technique is said to be robust if it is insensitive to deviations in the model and the input data from the ideal assumptions about them that were used in formulating the estimation process.
- Outlier
-
A data point with an error that does not belong to the assumed distribution of measurement errors.
- Lorentzian distribution
-
A distribution that resembles the normal distribution, but with lower probability for values that are close to the mean and higher probability for values that are farther from the mean.
- Linear
-
The models y = α1x + α2x2 and y = αexp(−x) are linear functions of the parameters α.
- Nonlinear
-
The models y = (α1x + α2x)2 and y = α1exp(−α2x) are nonlinear functions of the parameters α.
- Closed-form solution
-
A solution that can be expressed analytically in terms of a finite number of operations (for example, addition, multiplication, square root, and so on).
- Global optimization
-
The search for the lowest minimum or highest maximum of an objective function that has multiple minima or maxima. Such a function is called non-convex.
- Central limit theorem
-
The central limit theorem states that any variable that is calculated as the sum of a large number of variables, even if they are not normally distributed, will be normally distributed.
- Nonparametric methods
-
Statistical methods that do not assume an underlying distribution for the data being analysed.
- Number of degrees of freedom
-
The number of degrees of freedom in a regression is the number of data points that were used in the regression minus the number of estimated parameters.
- Null hypothesis
-
A statement that is tested for possible rejection under the assumption that it is true.
- Alternative hypothesis
-
A statement that is placed in opposition to the null hypothesis.
- Test-statistic
-
The variable calculated from the available data in order to test whether the null hypothesis can be rejected. Its distribution under the null hypothesis is usually known.
- Chi-square distribution
-
A variable that is calculated as the sum of the squares of ν variables that are N(0,1)-distributed has a Chi square (χ2)-distribution with ν degrees of freedom.
- P-value
-
The probability of obtaining a test-statistic at least as extreme as the one observed, assuming that the null hypothesis is true. It is effectively the probability of wrongly rejecting the null hypothesis when it is actually true.
- Significance value
-
The value below which a p-value supports rejecting the null hypothesis.
- F-distribution
-
A variable that is calculated as the ratio of two Chi-square-distributed variables divided by their degrees of freedom ν1 and ν2, has an F-distribution with ν1 and ν2 degrees of freedom.
- Trace
-
Sum of the diagonal elements of a matrix.
- Student's t-distribution
-
A distribution that is similar to N(0,1), except that it has heavier tails. It is a function of the number of degrees of freedom ν, and converges to N(0,1) as ν gets larger.
- Probabilistic process
-
A process in which the current state of a system does not uniquely determine its next state, but defines a set of possible states with their transition probabilities.
- Markov chain
-
A chain of events in which what happens at time point t + 1 only depends on what has happened at time point t, and not on any previous time points.
Rights and permissions
About this article
Cite this article
Jaqaman, K., Danuser, G. Linking data to models: data regression. Nat Rev Mol Cell Biol 7, 813–819 (2006). https://doi.org/10.1038/nrm2030
Published:
Issue Date:
DOI: https://doi.org/10.1038/nrm2030
This article is cited by
-
MLAGO: machine learning-aided global optimization for Michaelis constant estimation of kinetic modeling
BMC Bioinformatics (2022)
-
Simulation modelling for immunologists
Nature Reviews Immunology (2020)
-
Parameter estimation in models of biological oscillators: an automated regularised estimation approach
BMC Bioinformatics (2019)
-
Ranking network mechanisms by how they fit diverse experiments and deciding on E. coli's ammonium transport and assimilation network
npj Systems Biology and Applications (2019)
-
Glutamate spillover in C. elegans triggers repetitive behavior through presynaptic activation of MGL-2/mGluR5
Nature Communications (2019)