Combining metabolic modelling and machine learning for mechanism-aware predictions of cell behaviour

A fundamental tool for the inspection, interpretation, and exploitation of omic data is machine/deep learning, which has arguably fuelled several leaps forward in recent biomedical research, and is expected to increasingly drive it in the near future. In parallel, constraint-based modelling of metabolism is widely used due to its scope and flexibility, enabling mechanistic insights into the genotype–phenotype-environment relationship via integration with omic data.

These two computational frameworks have mostly been used in isolation, having distinct research communities associated with them. However, their complementary characteristics and common mathematical bases make them particularly suitable to be combined. I will describe how machine learning can be used with constraint-based modelling, discussing the mathematical and practical aspects involved. Compared with approaches applying machine learning to omic data directly, a multi-view approach merging experimentally and model-generated omic data can include key mechanistic information in an otherwise biology-agnostic learning process.

More on Claudio Angione can be found here.