An Introduction to Uncertainty in the Development of Computational Models of Biological Processes

Abstract

This chapter aims to provide an introduction to the different ways in which uncertainty can be dealt with computational modelling of biological processes. The first step is model establishment under uncertainty. Once models have been established, data can further be used to select which of the proposed models best meets the predefined criteria. Subsequently, parameter values can be optimized for a specific model configuration. Sensitivity analyses allow to assess the influence of the previous choices on the model output. Additionally, model adaptation permits to focus on specific aspects of the model without losing its global predictive capacity. Finally, predictions with the established models should also consider the effect of uncertainty in the model development process.

Simulation of live-cell imaging system reveals hidden uncertainties in cooperative binding measurements

“…The key to successful reporting of experimental results is to provide an objective evaluation and representation of the uncertainties that arise from imprecision and inaccuracies in the experimental processes. The study and estimation of the experimental uncertainties have been generally known as error analysis, its main function being to allow biophysicists to numerically indicate the validity and confidence of their experimental results [6][7][8].…”

“…A key insight into evaluating systematic variance is to estimate how well the ground-true model properties can be restored through analytical procedures, influencing the interpretation of biological properties reconstructed (or extracted) from actual biological images [6][7][8]. Such model-driven evaluation allows us to quantify the restoration efficiency and defects (or failure) in the reconstruction processes.…”

“…Systematic variance generally causes the reconstructed properties to be shifted in one direction from the ground-true model properties [6][7][8]. Of particular importance is the quantifying of systematic variance that arises from undetectable sources in the measurement processes.…”

“…Recent progress in robotic and automated techniques for sampling and analyzing complex biological data can reduce statistical uncertainties, increasing precisions in measuring biological and physical properties in living cells [1][2][3][4]. However, despite advances in computational techniques, the measurement process has resisted quantitative evaluation of systematic uncertainties that arise from inaccuracies in experimental data acquisition and analysis [5,6]. An absence and ignorance of evaluation of such systematic uncertainties often lead to excessive interpretations of the measured properties.…”

“…Although experienced experimental biophysicists are able to anticipate some systematic sources (e.g., the faulty calibration of measurement equipment) and ensure that most systematic uncertainties are much less than the required precision, other sources cannot even be detected by empirical approaches. The systematic variance that arises from undetectable sources in the measurement process can cause erroneous identification and interpretation of measured properties [5][6][7][8]. For example, structural uncertainties that arise from various model assumptions in biological network topologies cannot be directly extracted from experimental data, introducing errors into analysis and potentially giving rise to misleading conclusions [9,10].…”