Predicting microbial interactions with approaches based on flux balance analysis: an evaluation.

Area of Science:

• Microbial Ecology
• Systems Biology
• Computational Biology

Background:

• Genome-scale metabolic models (GEMs) and flux balance analysis (FBA) predict microbial growth and interactions.
• FBA is increasingly applied to microbial consortia for predicting interactions via in silico co-culture growth rates.
• A systematic evaluation of FBA's accuracy for predicting microbial interactions is lacking.

Purpose of the Study:

• To systematically evaluate the accuracy of FBA-based methods for predicting human and mouse gut bacterial interactions.
• To assess the reliability of current FBA tools and databases for microbial community modeling.
• To investigate the impact of GEM quality and tool settings on prediction accuracy.

Main Methods:

• Collected 26 semi-curated GEMs from the AGORA database and 4 curated GEMs.
• Evaluated three FBA tools (COMETS, Microbiome Modeling Toolbox, MICOM) using literature-derived in vitro growth data.
• Compared in silico predicted growth rates and interaction strengths (ratios) with experimental data under various conditions.

Main Results:

• FBA predictions using semi-curated GEMs showed poor correlation with experimental growth rates and interaction strengths.
• Accuracy was notably higher when using curated GEMs compared to semi-curated ones.
• Tool settings and media composition influenced prediction accuracy, but not sufficiently for semi-curated models.

Conclusions:

• FBA predictions of growth rates using semi-curated GEMs are not accurate enough for reliable prediction of microbial interaction strengths.
• Curated GEMs are essential for accurate in silico prediction of microbial community dynamics.
• Further development is needed to improve the reliability of FBA for microbial interaction prediction.