On the effects of alternative optima in context-specific metabolic model predictions

TL;DR

This paper investigates the impact of multiple equally valid solutions in context-specific metabolic models, emphasizing the importance of analyzing alternative optima for accurate biological predictions.

Contribution

It introduces computational methods to analyze alternative optima in context-specific metabolic models, highlighting their significance in biological interpretation.

Findings

Alternative optima are common in context-specific models.

Analyzing these optima improves prediction specificity.

Methods to reduce ambiguity in model predictions are proposed.

Abstract

Recent methodological developments have facilitated the integration of high-throughput data into genome-scale models to obtain context-specific metabolic reconstructions. A unique solution to this data integration problem often may not be guaranteed, leading to a multitude of context-specific predictions equally concordant with the integrated data. Yet, little attention has been paid to the alternative optima resulting from the integration of context-specific data. Here we present computational approaches to analyze alternative optima for different context-specific data integration instances. By using these approaches on metabolic reconstructions for the leaf of Arabidopsis thaliana and the human liver, we show that the analysis of alternative optima is key to adequately evaluating the specificity of the predictions in particular cellular contexts. While we provide several ways to reduce the ambiguity in the context-specific predictions, our findings indicate that the existence of alternative optimal solutions warrant caution in detailed context-specific analyses of metabolism.