Metagenomic estimation of absolute bacterial biomass in the mammalian gut through host-derived read normalization

Absolute bacterial biomass estimation in the human gut is crucial for understanding microbiome dynamics and host-microbe interactions. Current methods for quantifying bacterial biomass in stool, such as flow cytometry, quantitative polymerase chain reaction (qPCR), or spike-ins, can be labor-intensive, costly, and confounded by factors like water content, DNA extraction efficiency, PCR inhibitors, and other technical challenges that add bias and noise. We propose a simple, cost-effective approach that circumvents some of these technical challenges: directly estimating bacterial biomass from metagenomes using bacterial-to-host (B:H) read count ratios. We compared B:H ratios to the standard methods outlined above, demonstrating that B:H ratios are useful proxies for bacterial biomass in stool and possibly in other host-associated substrates. B:H ratios in stool were correlated with bacterial-to-diet (B:D) read count ratios, but B:D ratios exhibited a substantial number of outlier points. Host read depletion methods reduced the total number of human reads in a given sample, but B:H ratios were strongly correlated before and after host read depletion, indicating that host read depletion did not reduce the utility of B:H ratios. B:H ratios showed expected variation between health and disease states and were generally stable in healthy individuals over time. Finally, we showed how B:H and B:D ratios can be used to track antibiotic treatment response and recovery. B:H ratios offer a convenient alternative to other absolute biomass quantification methods, without the need for additional measurements, experimental design considerations, or machine learning, enabling robust absolute biomass estimates directly from stool metagenomic data.IMPORTANCEIn this study, we asked whether normalization by host reads alone was sufficient to estimate absolute bacterial biomass directly from stool metagenomic data, without the need for synthetic spike-ins, additional experimental biomass measurements, or training data. The approach assumes that the contribution of host DNA to stool is more constant or stable than biologically relevant fluctuations in bacterial biomass. We find that host read normalization is an effective method for detecting variation in gut bacterial biomass. Absolute bacterial biomass is a key metric that often gets left out of gut microbiome studies, and empowering researchers to include this measure more broadly in their metagenomic analyses should serve to improve our understanding of host-microbiota interactions.