Improved Metabolic Flux Estimations through Compositional Data Analysis

Isotopic Metabolic Flux Analysis (I-MFA) is a widely used method for estimating intracellular metabolic fluxes. The process involves comparing simulated and measured metabolite isotopologue distributions (MIDs) from isotope labeling experiments. MIDs are fractional abundances that always add up to one for a specific metabolite, making them compositional data.

The current state-of-the-art estimation methods, however, calculate standard Euclidean distances between MIDs in a non-compositional manner, leading to systematic bias.

The researchers in this study propose a new approach called compositional I-MFA to address this issue. They demonstrate how to create a meaningful orthonormal basis for MIDs using ordered sequential binary partitioning, which can then be transformed using isometric log-ratio (ILR) transformation. This modification allows them to estimate fluxes by minimizing Euclidean distances between ILR-transformed MIDs, making only a minor change to existing I-MFA workflows.

To test the effectiveness of their proposed framework, the researchers compared it against traditional methods using both a simplified model and a more complex, biologically realistic model. They evaluated the performance of the compositional I-MFA method in terms of point estimates, variability across different true fluxes, and confidence intervals.

The results showed that compositional I-MFA consistently outperformed traditional approaches, reducing the mean squared error of flux point estimates by an average of 42.6% and significantly narrowing confidence intervals in both examples.

In conclusion, the researchers argue that incorporating compositional data analysis into I-MFA significantly improves the accuracy of the method and can be implemented as a straightforward replacement for existing pipelines.