Understanding of the interrelationship between methane production and microorganisms in high-solid anaerobic co-digestion using microbial analysis and machine learning

Co-digestion of lignocellulosic biomass and animal manure is a common approach to increase the efficiency of methane production, but the niche differentiation and microbial metabolism of the anaerobic digestion (AD) process remain to be determined. To further explore the effect of the interaction between species and their compositional niches, the methane yield and resulting microbial community were determined by machine learning (ML) and 16S rRNA gene sequencing in mixed high-solid anaerobic digestion (HS-AD) with spray-enhanced conditions to explore the internal relationship between physical and chemical parameters and microorganisms and to speculate on the enhancement mechanism of co-digestion. In this study, three ML models (extreme learning machine (ELM), artificial neural network (ANN), and random forest (RF)) were applied to analyse and model AD of dry fermentation. The results showed that the best prediction model, based on ELM, could best predict the material biogas production in this experiment with a mean absolute error (MAE/10) of 0.678 and a coefficient of determination (R2) of 0.9574, whereas the characteristic percentage analysis of the RF model showed that butyric acid, acetic acid, and pH had three important influences on the biogas production values. Meanwhile, the results of high-throughput 16S rRNA gene sequencing and PICRUSt showed that the addition of manure containing ammonia nitrogen improved the metabolism of amino acids, enriched species capable of Clostridiales and Methanosarcinales, promoted the electronic transfer of nutrient metabolism, and thus enhanced AD. Moreover, the co-occurrence network showed that seven niches were differentiated within the HS-AD system to reduce the shock of ammonia nitrogen for methanogens. Overall, microbial analysis and ML can help understand the dynamic processes of methanogenic microorganisms and predict biogas production for the efficient operation of AD.