skip to content

Issues in Flux Balance Analysis

Presented by: 
D Fell Oxford Brookes University
Tuesday 4th November 2014 - 10:35 to 10:50
INI Seminar Room 1
Co-authors: Mark Poolman (Oxford Brookes University), Hassan Hartman (Oxford Brookes University)

As the practice of genome scale metabolic modelling by FBA has become more widespread, certain procedures and assumptions have been automatically adopted, almost as a standard, whereas their utility and applicability should be assessed for each specific model and investigation. In addition, there are some recurrent biochemical errors that are not always being filtered out.

Amongst the procedures not being given sufficient thought are: 1. Optimisation by maximisation or minimisation? Maximisation is generally adopted even though the way that the linear programming algorithm operates results in artefacts in the solutions that are not present on minimisation and that require post-processing to remove. 2. Expressing the biomass formation as a pseudo-reaction with non-integer stoichiometry or as part of the constraints in the analysis. Again the former is more general even though it is almost impossible to ensure it is correctly stoichiometrically balanced and feasible. 3. Substrate consumption for non-growth associated cell maintenance is for ATP generation. We have shown that part of this extra substrate consumption in Arabidopsis cells is for NADPH, presumably to combat oxidative damage, and this has consequences for the predicted fluxes in central carbon metabolism.

Recurrent errors include: 1. Writing enzyme prosthetic groups as substrates and products of enzyme reactions. FAD and FADH are the most frequent culprits. This creates pool metabolites that could generate spurious redox interactions across the network that will not exist because these groups are contained and recycled entirely within the enzyme reaction. 2. ATP from nothing. There are published models that can generate the ATP to satisfy maintenance requirements without any flows into the metabolic network from external material. Needless to say, all subsequent analysis of such a model is valueless. Preventing this is an elementary reality check during model construction.

The video for this talk should appear here if JavaScript is enabled.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.
University of Cambridge Research Councils UK
    Clay Mathematics Institute London Mathematical Society NM Rothschild and Sons