Comparing the performance of functional versus taxonomic metagenomics for detecting ammonia disturbances in the biogas system.

Biogas is a renewable energy source with great potential, but its production is frequently hindered by process disturbances, of which a high ammonia concentration is one common cause. It is desirable that such disturbances are found as early as possible; metagenomics data has the potential to improve this detection. This study compares functional and taxonomic aspects of metagenomics data, hypothesizing that functional data will perform better for detecting ammonia disturbances. The hypothesis was tested by metagenomic sequencing of samples from three independent studies, which followed lab-scale reactors during ammonia disturbances. The resulting sequences were used to predict genes, which were functionally and taxonomically annotated. The read counts of these features were fitted to disturbance states and ammonia concentrations of reactor samples using regularized regression, which allowed filtering out irrelevant features even with limited sample sizes. Within studies, taxonomic data had similar or better performance in detecting ammonia disturbances and in fitting ammonia concentrations. When applying trained models to other studies however, while performance was generally poor, functional models more often performed better compared to taxonomic models than the other way around. All in all, our hypothesis that functional metagenomics would outperform taxonomic metagenomics only found limited support.