puzzle. © shuttlecock

Modelling is everywhere

How is a model constructed?

The choice of model is determined by the data, the phenomenon being studied and the purpose of the research. A model is always designed jointly with researchers who are specialized in the field.

By Evelyne Lhoste - Pascale Mollier, translated by Emma Morton-Saliou
Updated on 05/23/2013
Published on 04/15/2013

Dessin humoristique Goff. © INRA
© INRA

The modelling process is the result of the interdisciplinary work of biologists, mathematicians and computer scientists. Depending on the specialty involved and the scope of the application, the biologists may be geneticists, cell biologists, epidemiologists or agronomists, to name a few.

A cyclical process

The modelling/experimentation cycle – The interpretation of results may reveal glitches that bring scientists back to the deployment phase to change the model. Meticulous interpretation can also uncover design flaws, in which case a new model is needed. If the interpretation phase yields insufficient sensitivity, the experimental design must be reconsidered. © INRA, Michaël Chelle. © INRA
The modelling/experimentation cycle – The interpretation of results may reveal glitches that bring scientists back to the deployment phase to change the model. Meticulous interpretation can also uncover design flaws, in which case a new model is needed. If the interpretation phase yields insufficient sensitivity, the experimental design must be reconsidered. © INRA, Michaël Chelle © INRA

Building a model involves going back and forth between multiple phases:
1. Identification of the exact goals of the model
2. Assessment of existing knowledge and collection of data on the biological phenomenon being studied
3. Choice of mathematical model and translation into computer code
4. Verification of the translation of the model into computer code
5. Analysis using mathematics and simulation of the sensitivity and characteristics of the model
6. Experiment planning: defining what data is needed, identifying collection methods and optimizing experimental protocol
7. Experimentation by biologist and verification of reliability of model
8. Identification and estimation of model’s parameters
9. Validation and definition of scope of the model

Importance of baseline data

Different kinds of data can be used in modelling, but usually numerical values collected from physical and/or biological measurements, biochemical analyses or imaging are used, as are documents written in plain language. Images are interpreted using various algorithms. Text mining programs translate data into formal language and classify it into databases.

Whatever the context, data must be collected carefully, sometimes over extended periods, and then pre-processed, sorted and collected from databases. The characteristics of data can be a factor in choosing theoretical models. Data can be wide-ranging or focus on a small group of individuals, and can be incomplete, censored or heterogeneous, in which case effective statistical methods are preferred over more direct, computer-generated algorithms.

Types of models

Empirical models are based on a statistical treatment of data. They are designed to replicate observed input/output relationships without deconstructing how they work. These models are also referred to as black-box models. Mechanistic or knowledge-based models, in contrast, describe the mechanisms and quantitative rules governing how a given system functions in more or less detail. Such models can be used to develop decision-making tools if they can predict how a system will function if parameters are changed.