Workshop title:  Predictive modelling methodologies and guidelines to data generators

The workshop was held on 1 July 2008 at the Crown Plaza Hotel near the airport, Dublin, Ireland. Panos Skandamis from the Agricultural University of Athens was the coordinator of this workshop.

The goal of the workshop is to increase the level of understanding of predictive modelling.

The presentation slides from this workshop can be found via the Members Only link. 

Background

Demonstration with methodology of development and validation of predictive models will be performed, including experimental design for collection of data for modelling microbial inactivation, probability of growth/no growth interface, classification of models (primary, secondary, tertiary), fitting primary and secondary models with linear and nonlinear regression, and validation of models by using performance statistics and visible comparisons of predicted values with observed values. The partners will receive training on the definition of dependent and independent variables as well as on the necessary data transformation in order to be further used for predictive modelling. Demonstrations will take place on recording kinetic or binary probability of growth data. There will also be a special session to demonstrate Monte Carlo simulation and how it may be implemented in exposure assessment.

Target Group

BIOTRACER partners involved in WPs which develop techniques and generate modelling data. Actually, all partners without experience or with limited knowledge on tracking models are essential to attend the workshop, in order to comprehend the type of data and experiments that are or will be requested by modellers.

Expected Outcome

Workshop participants will become familiar with the process of development and validation of predictive models. They will also understand what type of data may be used for modelling and how they can evaluate the suitability of their own data for modelling. The ultimate outcome will be to establish complete understanding between partners that produce data and modelers, in order the former to implement the requests by modelers, within the next phases of BIOTRACER.

Biotracer participants will pay the travel from their own BIOTRACER travel budget, while the meeting facilities will be paid by the BIOTRACER training budget.

Workshop Report from Panos Skandamis

Initially, a lecture was given on classification of tracking models and the methodology of model development and validation. During this lectures, extended discussions were performed in order to cover various questions of the attendees. In particular, most people were concerned on the criteria used to design an experiment for collection of modelling data (i.e., number of factors and levels per factor), as well as on statistical issues (e.g. minimum number of replicates and reliability of parameters in case of data gaps, etc.). People were also interesting in differentiating the difference between many replicates on a few single points and a few replicates on all data points comprising a growth or inactivation curve. It was made clear that independent experiments are important in order to address that the growth or inactivation trend under certain environmental conditions is replicable. A significant discussion also took place on how tracking models could be used in benefit of the food industry, especially under the concept of biotraceability.  

Following a break, the second part of the workshop started. During the session, Panos Skandamis delivered a folder with various XL files demonstrating fitting of individual growth/survival curves, mathematical expression of maximum specific growth rate as a function of temperature and pH, as well as explanation and manual estimation of statistical performance indices. Partners performed extended practice on the above issues, following solved and unsolved examples.  

Overall, it seemed that partners obtained a spherical understanding on the concepts of tracking models and the statistical qualifications required to evaluate the performance of these models. 

Agenda

13.00-13.30    Classification of predictive models.
13.30-14.15    Methodology for development and validation of predictive models.

  • Experimental design
  • Data collection
  • Fitting procedures
  • Simulation procedures

14.15-15.00    Reviews on tracking models within the scope of BIOTRACER.
15.00-15.30    Coffee break and demonstration of predictive modelling software (Growth Predictor, Pathogen Modelling Program, Combase toolbox and Symprevius).
15.30-16.30    Demos on the fitting of primary and secondary models to the data produced in BIOTRACER.

16.30-17.30     Practical exercises with prototype and participant-provided data.

More photos