What machine learning contributes

Machine learning uses data to learn relationships that can be useful for prediction, classification or representation. Mathematical biology often begins instead with equations describing biological mechanisms.

Complementary roles

A mechanistic model asks how specified biological processes generate observed behaviour. Machine learning can discover predictive patterns in large or complex data without requiring every relationship to be written explicitly in advance.

ML can assist with parameter estimation, surrogate modelling, forecasting, image or sequence classification and discovering relationships that may later motivate mechanistic hypotheses.

Limits

Predictive accuracy does not automatically establish a biological mechanism or causal relationship. Performance also depends strongly on whether training data represent the setting in which the model is used.