An answer to the question: How do you know your model is correct?

The modeling process is just quantitative hypothesis testing. We have formulated a hypothesis, converted it to the corresponding system of equations, and tested it against the experimental data. What we can say is that the model is quantitatively consistent with the data.

No model can be said to be "correct", but some models can be shown to be consistent and
some inconsistent with the data. Inconsistent models can be rejected; consistent models yield useful information and are worthy of further experimental testing.

We can easily test any alternative hypothesis that someone would care to propose. But you should remind them of the enormous volume of data that an alternative model would have to account for; you can actually list the data sets you have fitted and the constraints you have obeyed.

Modeling simply quantifies the comparison of theory with experiment. Modeling is thus a useful tool in applying traditional scientific method, but it has all of the well-known limitations of scientific method. In particular, no theory is ever proven correct. This is because it may not account for data from future experiments. Modeling is powerful, but it is not magic. When you present a model, you must remain a scientist. Don't let the audience hold you to a higher standard than they apply to their own work. Can you imagine what would happen if a questioner asked at the end of a talk: "How do you know your explanation is correct as opposed to several alternative explanations?" Unless the presenter has already tested the proposed alternatives and found them inconsistent, it would have to be admitted that the alternative explanations are possible. No scientist is ever certain that his or her explanation or model or theory is correct.

There is more detail on these issues in my chapter on the Rationale for Kinetic Modeling

Return to Computational Biology FAQ