Robustness Analysis of Stochastic Biochemical Systems

We propose a new framework for rigorous robustness analysis of stochastic biochemical systems that is based on probabilistic model checking techniques. We adapt the general definition of robustness introduced by Kitano to the class of stochastic systems modelled as continuous time Markov Chains in order to extensively analyse and compare robustness of biological models with uncertain parameters. The framework utilises novel computational methods that enable to effectively evaluate the robustness of models with respect to quantitative temporal properties and parameters such as reaction rate constants and initial conditions. We have applied the framework to gene regulation as an example of a central biological mechanism where intrinsic and extrinsic stochasticity plays crucial role due to low numbers of DNA and RNA molecules. Using our methods we have obtained a comprehensive and precise analysis of stochastic dynamics under parameter uncertainty. Furthermore, we apply our framework to compare several variants of two-component signalling networks from the perspective of robustness with respect to intrinsic noise caused by low populations of signalling components. We have successfully extended previous studies performed on deterministic models (ODE) and showed that stochasticity may significantly affect obtained predictions. Our case studies demonstrate that the framework can provide deeper insight into the role of key parameters in maintaining the system functionality and thus it significantly contributes to formal methods in computational systems biology.     

Application of dynamic metabolic flux analysis for process modeling: Robust flux estimation with regularization,confidence bounds,and selection of elementary modes

In macroscopic dynamic models of fermentation processes, elementary modes (EM) derived from metabolic networks are often used to describe the reaction stoichiometry in a simplified manner and to build predictive models by parameterizing kinetic rate equations for the EM. In this procedure, the selection of a set of EM is a key step which is followed by an estimation of their reaction rates and of the associated confidence bounds. In this paper, we present a method for the computation of reaction rates of cellular reactions and EM as well as an algorithm for the selection of EM for process modeling. The method is based on the dynamic metabolic flux analysis (DMFA) proposed by Leighty and Antoniewicz (2011, Metab Eng, 13(6), 745–755) with additional constraints, regularization and analysis of uncertainty. Instead of using estimated uptake or secretion rates, concentration measurements are used directly to avoid an amplification of measurement errors by numerical differentiation. It is shown that the regularized DMFA for EM method is significantly more robust against measurement noise than methods using estimated rates. The confidence intervals for the estimated reaction rates are obtained by bootstrapping. For the selection of a set of EM for a given st oichiometric model, the DMFA for EM method is combined with a multiobjective genetic algorithm. The method is applied to real data from a CHO fed-batch process. From measurements of six fed-batch experiments, 10 EM were identified as the smallest subset of EM based upon which the data can be described sufficiently accurately by a dynamic model. The estimated EM reaction rates and their confidence intervals at different process conditions provide useful information for the kinetic modeling and subsequent process optimization.     

Prediction of non-linear pharmacokinetics in humans of an antibody-drug conjugate (ADC) when evaluation of higher doses in animals is limited by tolerability: Case study with an anti-CD33 ADC

For antibody-drug conjugates (ADCs) that carry a cytotoxic drug, doses that can be administered in preclinical studies are typically limited by tolerability, leading to a narrow dose range that can be tested. For molecules with non-linear pharmacokinetics (PK), this limited dose range may be insufficient to fully characterize the PK of the ADC and limits translation to humans. Mathematical PK models are frequently used for molecule selection during preclinical drug development and for translational predictions to guide clinical study design. Here, we present a practical approach that uses limited PK and receptor occupancy (RO) data of the corresponding unconjugated antibody to predict ADC PK when conjugation does not alter the non-specific clearance or the antibody-target interaction. We used a 2-compartment model incorporating non-specific and specific (target mediated) clearances, where the latter is a function of RO, to describe the PK of anti-CD33 ADC with dose-limiting neutropenia in cynomolgus monkeys. We tested our model by comparing PK predictions based on the unconjugated antibody to observed ADC PK data that was not utilized for model development. Prospective prediction of human PK was performed by incorporating in vitro binding affinity differences between species for varying levels of CD33 target expression. Additionally, this approach was used to predict human PK of other previously tested anti-CD33 molecules with published clinical data. The findings showed that, for a cytotoxic ADC with non-linear PK and limited preclinical PK data, incorporating RO in the PK model and using data from the corresponding unconjugated antibody at higher doses allowed the identification of parameters to characterize monkey PK and enabled human PK predictions.     

Spatial Uncertainty and Ecological Models

Applied ecological models that are used to understand and manage natural systems often rely on spatial data as input. Spatial uncertainty in these data can propagate into model predictions. Uncertainty analysis, sensitivity analysis, error analysis, error budget analysis, spatial decision analysis, and hypothesis testing using neutral models are all techniques designed to explore the relationship between variation in model inputs and variation in model predictions. Although similar methods can be used to answer them, these approaches address different questions. These approaches differ in (a) whether the focus is forward or backward (forward to evaluate the magnitude of variation in model predictions propagated or backward to rank input parameters by their influence); (b) whether the question involves model robustness to large variations in spatial pattern or to small deviations from a reference map; and (c) whether processes that generate input uncertainty (for example, cartographic error) are of interest. In this commentary, we propose a taxonomy of approaches, all of which clarify the relationship between spatial uncertainty and the predictions of ecological models. We describe existing techniques and indicate a few areas where research is needed.     

Uncertainty in predictions of range dynamics: black grouse climbing the Swiss Alps

Empirical species distribution models (SDMs) constitute often the tool of choice for the assessment of rapid climate change effects on species’ vulnerability. Conclusions regarding extinction risks might be misleading, however, because SDMs do not explicitly incorporate dispersal or other demographic processes. Here, we supplement SDMs with a dynamic population model 1) to predict climate‐induced range dynamics for black grouse in Switzerland, 2) to compare direct and indirect measures of extinction risks, and 3) to quantify uncertainty in predictions as well as the sources of that uncertainty. To this end, we linked models of habitat suitability to a spatially explicit, individual‐based model. In an extensive sensitivity analysis, we quantified uncertainty in various model outputs introduced by different SDM algorithms, by different climate scenarios and by demographic model parameters. Potentially suitable habitats were predicted to shift uphill and eastwards. By the end of the 21st century, abrupt habitat losses were predicted in the western Prealps for some climate scenarios. In contrast, population size and occupied area were primarily controlled by currently negative population growth and gradually declined from the beginning of the century across all climate scenarios and SDM algorithms. However, predictions of population dynamic features were highly variable across simulations. Results indicate that inferring extinction probabilities simply from the quantity of suitable habitat may underestimate extinction risks because this may ignore important interactions between life history traits and available habitat. Also, in dynamic range predictions uncertainty in SDM algorithms and climate scenarios can become secondary to uncertainty in dynamic model components. Our study emphasises the need for principal evaluation tools like sensitivity analysis in order to assess uncertainty and robustness in dynamic range predictions. A more direct benefit of such robustness analysis is an improved mechanistic understanding of dynamic species’ responses to climate change.     

A kinetic model of Escherichia coli core metabolism satisfying multiple sets of mutant flux data

In contrast to stoichiometric-based models, the development of large-scale kinetic models of metabolism has been hindered by the challenge of identifying kinetic parameter values and kinetic rate laws applicable to a wide range of environmental and/or genetic perturbations. The recently introduced ensemble modeling (EM) procedure provides a promising remedy to address these challenges by decomposing metabolic reactions into elementary reaction steps and incorporating all phenotypic observations, upon perturbation, in its model parameterization scheme. Here, we present a kinetic model of Escherichia coli core metabolism that satisfies the fluxomic data for wild-type and seven mutant strains by making use of the EM concepts. This model encompasses 138 reactions, 93 metabolites and 60 substrate-level regulatory interactions accounting for glycolysis/gluconeogenesis, pentose phosphate pathway, TCA cycle, major pyruvate metabolism, anaplerotic reactions and a number of reactions in other parts of the metabolism. Parameterization is performed using a formal optimization approach that minimizes the discrepancies between model predictions and flux measurements. The predicted fluxes by the model are within the uncertainty range of experimental flux data for 78% of the reactions (with measured fluxes) for both the wild-type and seven mutant strains. The remaining flux predictions are mostly within three standard deviations of reported ranges. Converting the EM-based parameters into a Michaelis–Menten equivalent formalism revealed that 35% of K_m and 77% of k_cat parameters are within uncertainty range of the literature-reported values. The predicted metabolite concentrations by the model are also within uncertainty ranges of metabolomic data for 68% of the metabolites. A leave-one-out cross-validation test to evaluate the flux prediction performance of the model showed that metabolic fluxes for the mutants located in the proximity of mutations used for training the model can be predicted more accurately. The constructed model and the parameterization procedure presented in this study pave the way for the construction of larger-scale kinetic models with more narrowly distributed parameter values as new metabolomic/fluxomic data sets are becoming available for E. coli and other organisms.     

Hybrid physics-based and data-driven modeling for bioprocess online simulation and optimization

Model-based online optimization has not been widely applied to bioprocesses due to the challenges of modeling complex biological behaviors, low-quality industrial measurements, and lack of visualization techniques for ongoing processes. This study proposes an innovative hybrid modeling framework which takes advantages of both physics-based and data-driven modeling for bioprocess online monitoring, prediction, and optimization. The framework initially generates high-quality data by correcting raw process measurements via a physics-based noise filter (a generally available simple kinetic model with high fitting but low predictive performance); then constructs a predictive data-driven model to identify optimal control actions and predict discrete future bioprocess behaviors. Continuous future process trajectories are subsequently visualized by re-fitting the simple kinetic model (soft sensor) using the data-driven model predicted discrete future data points, enabling the accurate monitoring of ongoing processes at any operating time. This framework was tested to maximize fed-batch microalgal lutein production by combining with different online optimization schemes and compared against the conventional open-loop optimization technique. The optimal results using the proposed framework were found to be comparable to the theoretically best production, demonstrating its high predictive and flexible capabilities as well as its potential for industrial application.     

K-FIT: An accelerated kinetic parameterization algorithm using steady-state fluxomic data

Kinetic models predict the metabolic flows by directly linking metabolite concentrations and enzyme levels to reaction fluxes. Robust parameterization of organism-level kinetic models that faithfully reproduce the effect of different genetic or environmental perturbations remains an open challenge due to the intractability of existing algorithms. This paper introduces Kinetics-based Fluxomics Integration Tool (K-FIT), a robust kinetic parameterization workflow that leverages a novel decomposition approach to identify steady-state fluxes in response to genetic perturbations followed by a gradient-based update of kinetic parameters until predictions simultaneously agree with the fluxomic data in all perturbed metabolic networks. The applicability of K-FIT to large-scale models is demonstrated by parameterizing an expanded kinetic model for E. coli (307 reactions and 258 metabolites) using fluxomic data from six mutants. The achieved thousand-fold speed-up afforded by K-FIT over meta-heuristic approaches is transformational enabling follow-up robustness of inference analyses and optimal design of experiments to inform metabolic engineering strategies.     

Towards kinetic modeling of genome-scale metabolic networks without sacrificing stoichiometric,thermodynamic and physiological constraints

Mathematical modeling is an essential tool for the comprehensive understanding of cell metabolism and its interactions with the environmental and process conditions. Recent developments in the construction and analysis of stoichiometric models made it possible to define limits on steady-state metabolic behavior using flux balance analysis. However, detailed information on enzyme kinetics and enzyme regulation is needed to formulate kinetic models that can accurately capture the dynamic metabolic responses. The use of mechanistic enzyme kinetics is a difficult task due to uncertainty in the kinetic properties of enzymes. Therefore, the majority of recent works considered only mass action kinetics for reactions in metabolic networks. Herein, we applied the optimization and risk analysis of complex living entities (ORACLE) framework and constructed a large-scale mechanistic kinetic model of optimally grown Escherichia coli. We investigated the complex interplay between stoichiometry, thermodynamics, and kinetics in determining the flexibility and capabilities of metabolism. Our results indicate that enzyme saturation is a necessary consideration in modeling metabolic networks and it extends the feasible ranges of metabolic fluxes and metabolite concentrations. Our results further suggest that enzymes in metabolic networks have evolved to function at different saturation states to ensure greater flexibility and robustness of cellular metabolism.     

State and parameter estimation of microalgal photobioreactor cultures based on local irradiance measurement

Local photosynthetic photon flux fluence rate (PPFFR) determined by a submersible 4pi quantum micro-sensor was used in developing a versatile on-line state estimator for stirred-tank microalgal photobioreactor cultures. A marine micro-alga Dunaliella salina was used as a model organism in this study. On-line state estimation was realized using the extended Kalman filter (EKF), based on a state model of the photobioreactor and on-line local PPFFR measurement. The dynamic state model for the photobioreactor was derived based on mass-balance equations of the relevant states. The measurement equation was established based on an empirical correlation between the microalgal biomass concentration and the local PPFFR measured at a fixed point inside the photobioreactor. An internal model approach was used to estimate the specific growth rate without the need of state-based kinetic expression. The estimator was proven to be capable of estimating biomass concentration and specific growth rate, as well as phosphate and dissolved oxygen concentrations in a photobioreactor illuminated with either fixed or time-varying incident radiation. The quantum sensor was shown to be robust and able to quickly respond to dynamic changes in local PPFFR. In addition, the quantum sensor outputs were not affected by bubble aeration or agitation within the typical operating range. The strong filtering capacity of EKF gives the state estimator superior performance compared to direct calculation from the empirical biomass/local PPFFR correlation. This state estimation system makes use of inexpensive and reliable sensor hardware to report key process dynamics of microalgal photobioreactor cultures on-line, enabling improved operation of such a process.     

Site-specific antibody-drug conjugate heterogeneity characterization and heterogeneity root cause analysis

ABSTRACT

Site-specific antibody-drug conjugates (ADCs) are designed to overcome the heterogeneity observed with first-generation ADCs that use random conjugation to surface-exposed lysine residues or conjugation to interchain disulfide bonds. Despite significantly enhanced homogeneity, however, the production of site-specific ADCs yields some process-related species heterogeneity, including stereoisomers, unconjugated antibody, underconjugated species, and overconjugated species. An elevated level of size variants, such as heavy chain-light chain species (half ADC), heavy chain-heavy chain-light chain species, and light chain species, is also observed with the final site-specific ADC product. To understand the root cause of heterogeneity generated during the ADC conjugation process, we designed time-course studies for each conjugation step, including reduction, oxidation, conjugation, and quenching. We developed both non-reduced peptide map and LabChip-based capillary electrophoresis sodium dodecyl sulfate methods for time-course sample analysis. On the basis of our time-course data, the half ADC and unconjugated antibody were generated during oxidation as a result of alternative disulfide bond arrangements. During oxidation, two hinge cysteines formed an intra-chain disulfide bond in the half ADC, and three inter-chain hinge disulfide bonds were formed in the unconjugated antibody. Time-course data also showed that the elevated level of size variants, especially heavy chain-heavy chain-light chain species and light chain species, resulted from the quenching step, where the quenching reagent engaged in a disulfide bond exchange reaction with the ADC and broke the disulfide bonds connecting the heavy chain and light chain. Underconjugated and overconjugated species arose from the equilibrium established during the conjugation reaction.     

Modeling changes in biomass composition during microwave‐based alkali pretreatment of switchgrass

This study used two different approaches to model changes in biomass composition during microwave‐based pretreatment of switchgrass: kinetic modeling using a time‐dependent rate coefficient, and a Mamdani‐type fuzzy inference system. In both modeling approaches, the dielectric loss tangent of the alkali reagent and pretreatment time were used as predictors for changes in amounts of lignin, cellulose, and xylan during the pretreatment. Training and testing data sets for development and validation of the models were obtained from pretreatment experiments conducted using 1–3% w/v NaOH (sodium hydroxide) and pretreatment times ranging from 5 to 20 min. The kinetic modeling approach for lignin and xylan gave comparable results for training and testing data sets, and the differences between the predictions and experimental values were within 2%. The kinetic modeling approach for cellulose was not as effective, and the differences were within 5–7%. The time‐dependent rate coefficients of the kinetic models estimated from experimental data were consistent with the heterogeneity of individual biomass components. The Mamdani‐type fuzzy inference was shown to be an effective approach to model the pretreatment process and yielded predictions with less than 2% deviation from the experimental values for lignin and with less than 3% deviation from the experimental values for cellulose and xylan. The entropies of the fuzzy outputs from the Mamdani‐type fuzzy inference system were calculated to quantify the uncertainty associated with the predictions. Results indicate that there is no significant difference between the entropies associated with the predictions for lignin, cellulose, and xylan. It is anticipated that these models could be used in process simulations of bioethanol production from lignocellulosic materials. Biotechnol. Bioeng. 2010;105: 88–97. © 2009 Wiley Periodicals, Inc.     

Process analytical technology case study: Part II. Development and validation of quantitative near-infrared calibrations in support of a process analytical technology application for real-time release

This article is the second of a series of articles detailing the development of near-infrared (NIR) methods for solid dosage-form analysis. Experiments were conducted at the Duquesne University Center for Pharmaceutical Technology to demonstrate a method for developing and validating NIR models for the analysis of active pharmaceutical ingredient (API) content and hardness of a solid dosage form. Robustness and cross-validation testing were used to optimize the API content and hardness models. For the API content calibration, the optimal model was determined as multiplicative scatter correction with Savitsky-Golay first-derivative preprocessing followed by partial least-squares (PLS) regression including 4 latent variables. API content calibration achieved root mean squared error (RMSE) and root mean square error of cross validation (RMSECV) of 1.48 and 1.80 mg, respectively. PLS regression and baseline-fit calibration models were compared for the prediction of tablet hardness. Based on robustness testing, PLS regression was selected for the final hardness model, with RMSE and RMSECV of 8.1 and 8.8 N, respectively. Validation testing indicated that API content and hardness of production-scale tablets is predicted with root mean square error of prediction of 1.04 mg and 8.5 N, respectively. Explicit robustness testing for high-flux noise and wavelength uncertainty demonstrated the robustness of the API concentration calibration model with respect to normal instrument operating conditions. Published: October 6, 2005 The views presented in this article do not necessarily reflect those of the Food and Drug Administration.     

农业废弃物养分循环利用技术模式评估模型的研究进展

农业废弃物的养分循环利用技术模式是实现农业循环经济的重要手段,其评估模型为优化养分循环利用技术提供了重要支撑。本文总结了农业废弃物养分循环技术模式评估框架、评估模型及评价指标、模型的数据源及其不确定性分析,以及模型应用尺度的研究进展。当前,常用于评估养分流动的模型主要是过程数学模型和产业生态学模型。过程数学模型和产业生态学模型在评估结果的可靠性和模拟尺度上存在较大差异,前者主要集中在实验室或中试规模,精度较高;后者可以实现从微观到宏观的多尺度模拟,数据的获取方式导致其具有较高的不确定性。最后,本文对农业废弃物养分循环利用技术评估模型的研究进行展望,提出为了在区域尺度上实现对农业生产系统废弃物资源化利用技术的准确评估,可以将过程数学模型与工业生态学模型相结合,建立可靠的模型框架和数据库,同时,在工厂、农场、村落、乡镇、区域等地理尺度进行模型拓展研究。     

Incorporating uncertainty in predictive species distribution modelling

Motivated by the need to solve ecological problems (climate change, habitat fragmentation and biological invasions), there has been increasing interest in species distribution models (SDMs). Predictions from these models inform conservation policy, invasive species management and disease-control measures. However, predictions are subject to uncertainty, the degree and source of which is often unrecognized. Here, we review the SDM literature in the context of uncertainty, focusing on three main classes of SDM: niche-based models, demographic models and process-based models. We identify sources of uncertainty for each class and discuss how uncertainty can be minimized or included in the modelling process to give realistic measures of confidence around predictions. Because this has typically not been performed, we conclude that uncertainty in SDMs has often been underestimated and a false precision assigned to predictions of geographical distribution. We identify areas where development of new statistical tools will improve predictions from distribution models, notably the development of hierarchical models that link different types of distribution model and their attendant uncertainties across spatial scales. Finally, we discuss the need to develop more defensible methods for assessing predictive performance, quantifying model goodness-of-fit and for assessing the significance of model covariates.     

The dynamical consequences of developmental variability and demographic stochasticity for host-parasitoid interactions

Few age-structured models of species dynamics incorporate variability and uncertainty in population processes. Motivated by laboratory data for an insect and its parasitoid, we investigate whether such assumptions are appropriate when considering the population dynamics of a single species and its interaction with a natural enemy. Specifically, we examine the effects of developmental variability and demographic stochasticity on different types of cyclic dynamics predicted by traditional models. We show that predictions based on the deterministic fixed-development approach are differentially sensitive to variability and noise in key life stages. In particular, we find that the demonstration of half-generation cycles in the single-species model and the multigeneration cycles in the host-parasitoid model are sensitive to the introduction of developmental variability and noise, whereas generation cycles are robust to the intrinsic variability and uncertainty that may be found in nature.     

Quasi-multiparameter sensitivity measure for robustness analysis of complex biochemical networks

Robustness is the ability to resume reliable operation in the face of different types of perturbations. Analysis of how network structure achieves robustness enables one to understand and design cellular systems. It is typically true that all parameters simultaneously differ from their nominal values in vivo, but there have been few intelligible measures to estimate the robustness of a system's function to the uncertainty of all parameters.We propose a numerical and fast measure of a robust property to the uncertainty of all kinetic parameters, named quasi-multiparameter sensitivity (QMPS), which is defined as the sum of the squared magnitudes of single-parameter sensitivities. Despite its plain idea, it has hardly been employed in analysis of biological models. While QMPS is theoretically derived as a linear model, QMPS can be consistent with the expected variance simulated by the widely used Monte Carlo method in nonlinear biological models, when relatively small perturbations are given. To demonstrate the feasibility of QMPS, it is employed for numerical comparison to analyze the mechanism of how specific regulations generate robustness in typical biological models.QMPS characterizes the robustness much faster than the Monte Carlo method, thereby enabling the extensive search of a large parameter space to perform the numerical comparison between alternative or competing models. It provides a theoretical or quantitative insight to an understanding of how specific network structures are related to robustness. In circadian oscillators, a negative feedback loop with multiple phosphorylations is demonstrated to play a critical role in generating robust cycles to the uncertainty of multiple parameters.     

Mathematical modeling and optimization of cellulase protein production using Trichoderma reesei RL-P37

The enzyme cellulase, a multienzyme complex made up of several proteins, catalyzes the conversion of cellulose to glucose in an enzymatic hydrolysis-based biomass-to-ethanol process. Production of cellulase enzyme proteins in large quantities using the fungus Trichoderma reesei requires understanding the dynamics of growth and enzyme production. The method of neural network parameter function modeling, which combines the approximation capabilities of neural networks with fundamental process knowledge, is utilized to develop a mathematical model of this dynamic system. In addition, kinetic models are also developed. Laboratory data from bench-scale fermentations involving growth and protein production by T. reesei on lactose and xylose are used to estimate the parameters in these models. The relative performances of the various models and the results of optimizing these models on two different performance measures are presented. An approximately 33% lower root-mean-squared error (RMSE) in protein predictions and about 40% lower total RMSE is obtained with the neural network-based model as opposed to kinetic models. Using the neural network-based model, the RMSE in predicting optimal conditions for two performance indices, is about 67% and 40% lower, respectively, when compared with the kinetic models. Thus, both model predictions and optimization results from the neural network-based model are found to be closer to the experimental data than the kinetic models developed in this work. It is shown that the neural network parameter function modeling method can be useful as a "macromodeling" technique to rapidly develop dynamic models of a process.     

Dynamic activation model for a glutamatergic neurovascular unit

This article considers a dynamic spatially lumped model for brain energy metabolism and proposes to use the results of a Markov chain Monte Carlo (MCMC) based flux balance analysis to estimate the kinetic model parameters. By treating steady state reaction fluxes and transport rates as random variables we are able to propagate the uncertainty in the steady state configurations to the predictions of the dynamic model, whose responses are no longer individual but ensembles of time courses. The kinetic model consists of five compartments and is governed by kinetic mass balance equations with Michaelis-Menten type expressions for reaction rates and transports between the compartments. The neuronal activation is implemented in terms of the effect of neuronal activity on parameters controlling the blood flow and neurotransmitter transport, and a feedback mechanism coupling the glutamate concentration in the synaptic cleft and the ATP hydrolysis, thus accounting for the energetic cost of the membrane potential restoration in the postsynaptic neurons. The changes in capillary volume follow the balloon model developed for BOLD MRI. The model follows the time course of the saturation levels of the blood hemoglobin, which link metabolism and BOLD FMRI signal. Analysis of the model predictions suggest that stoichiometry alone is not enough to determine glucose partitioning between neuron and astrocyte. Lactate exchange between neuron and astrocyte is supported by the model predictions, but the uncertainty on the direction and rate is rather elevated. By and large, the model suggests that astrocyte produces and effluxes lactate, while neuron may switch from using to producing lactate. The level of ATP hydrolysis in astrocyte is substantially higher than strictly required for neurotransmitter cycling, in agreement with the literature.