Identifying keystone species in the microbial community based on cross-sectional data

Background: In microbial communities, the keystone species have a greater impact on the performance and dynamics of ecosystem than that of other species, in which we can see from the results that losing gut microbiome causes some specific diseases. A number of ongoing studies aim at identifying links between microbial community structure and human diseases. Method: In this paper, we have introduced a valid keystone species identification method, in which a new Spread Intensity (SI) algorithm is used. Because the accuracies of current keystone species identification algorithms are difficult to evaluate for the high diversity and uncultivated status of microbial communities, we simulated cross-sectional data of microbial communities with known interactions and set up standard keystoneness rankings using Generalized Lotka-Volterra (GLV) model. Subsequently, we compared the SI algorithm with existing methods by using simulated data and obtained an obvious better performance of SI algorithm than other methods. Also, we applied this method to gut microbiota datasets and identified some microbes having the potential association with body weight. We first assembled three correlation metrics to calculate the interspecies correlation. Then we applied network deconvolution to remove indirect correlations. Finally, we used Molecular Ecological Network Analysis (MENA) to construct the co-occurrence network. According to experimental results, SI algorithm has an excellent performance in identifying highly correlated species in gut microbiome to body weight. Result: This result provides an effective indicator for modulating gut microbiota and thus enables the gene therapy and other gene-level treatments for losing-weight and other gut-associated diseases.