4.1 Approaches for CBM and ML integration
The integration of ML and CBM can be conducted in three chief ways
[26]: (a) The output of CBM is the flux distribution that indicates
the metabolic state of the cell. This fluxomic data can be trained
directly through ML methods to obtain more biological insight into the
desired system (Figure 5A ). (b) ML is an effective tool for
merging and analyzing heterogeneous omics datasets beyond ML
applications to single omics. By combining these multi-omics datasets
with GEMs, context-specific models are generated. More accurate flux
values obtained from context-specific GEMs can be re-integrated with
experimental omics data for further predictions (Figure 5B ).
(c) CBM models and fluxomic data can be produced directly by introducing
omics or multi-omics datasets into ML algorithms. All of the three
mentioned methods might be operated by supervised or unsupervised
algorithms (Figure 5C ).