14.9 Models as Biological Hypotheses

At this point, we have developed all the essential ingredients of a dynamic model. We can identify the relevant state variables, formulate differential equations describing their interactions, and simulate their behaviour over time.

But an important question remains:

Why do we build mathematical models at all?

One possible answer is that models allow us to predict experimental observations. While this is certainly true, prediction alone is not the primary goal of systems biology. A model that reproduces experimental data is useful only if it also helps us understand the biological mechanisms responsible for those observations.

Models formalize biological knowledge

Every mathematical model begins with a biological idea.

We may hypothesize that a protein activates gene expression, that an enzyme catalyses a metabolic reaction, or that a signalling molecule inhibits the activity of another protein. Initially, these ideas are qualitative descriptions of biological processes.

A mathematical model transforms these qualitative ideas into a quantitative framework. Every variable represents a measurable biological quantity, every parameter corresponds to a biological process, and every equation expresses an explicit assumption about how different components interact.

Constructing a model therefore forces us to state our biological assumptions precisely.

Models make hypotheses testable

Scientific hypotheses are valuable only if they can be tested experimentally.

Mathematical models provide exactly this opportunity. Once a biological hypothesis has been translated into a quantitative model, we can simulate its behaviour and compare the predictions with experimental observations.

If the predicted behaviour agrees with the data, the model supports the underlying biological hypothesis. If the predictions differ substantially from experimental measurements, the discrepancy indicates that our current understanding of the biological system is incomplete.

The disagreement itself is often more informative than the agreement. It points directly towards biological processes that have not yet been considered.

Every model is a simplification

No mathematical model captures every detail of a living system.

Cells contain thousands of interacting molecules, many of which influence one another in subtle and incompletely understood ways. Including every known process would produce models that are almost impossible to analyse or interpret.

Instead, models deliberately focus on the biological mechanisms that are believed to be most relevant for answering a particular question.

A successful model is therefore not necessarily the most detailed one. Rather, it is the simplest model that explains the experimental observations sufficiently well.

This principle, often referred to as parsimony, is fundamental throughout systems biology. By reducing biological complexity to its essential mechanisms, models help reveal general organizing principles that would otherwise remain hidden.

From prediction to understanding

The true value of a mathematical model lies in the biological insight it provides.

Once a model has been validated, it can be used to explore scenarios that are difficult or impossible to investigate experimentally. Individual parameters can be varied systematically, regulatory interactions can be removed or strengthened, and hypothetical perturbations can be performed without affecting a living organism.

In this way, models become virtual laboratories.

More importantly, they allow us to distinguish between correlation and mechanism. Experimental observations often reveal that two biological quantities change together. A mechanistic model asks a much deeper question: which biological processes generate this behaviour?

The iterative cycle of modelling

Mathematical modelling is therefore not a one-time exercise but an iterative scientific process.

The cycle typically begins with biological observations, from which a mechanistic hypothesis is formulated. This hypothesis is translated into a mathematical model and used to generate predictions. Experimental data are then collected to evaluate these predictions. The results either support the current model or reveal discrepancies that motivate its refinement.

This continuous interaction between experiments and modelling gradually improves our understanding of the biological system.

Rather than replacing experiments, mathematical models guide experiments towards the most informative questions.

Key concepts

Mathematical models formalize biological hypotheses.
Every variable, parameter, and equation has a biological interpretation.
Models generate quantitative predictions that can be tested experimentally.
Disagreement between model predictions and experimental data often reveals missing biological mechanisms.
Modelling is an iterative process that combines experimentation and theory.

Summary

Mathematical models are much more than computational tools. They provide a formal language for expressing biological hypotheses and testing mechanistic explanations. By comparing model predictions with experimental observations, researchers can evaluate competing hypotheses, identify missing mechanisms, and gradually improve their understanding of biological systems. In systems biology, the ultimate goal of modelling is therefore not simply prediction, but biological understanding.

Self-check questions

Why is the primary purpose of a mathematical model not simply to fit experimental data?
How does constructing a mathematical model make biological hypotheses testable?
Why are all mathematical models necessarily simplifications of reality?
What is meant by the principle of parsimony in modelling?
Why is modelling considered an iterative scientific process?