Design And Implementation Of Microbial Ecological Network Analysis System

Various microbes co-exist in microbial communities,forming complicated network through multiple interactions.They are closely related to industrial production,living things,and even cast direct impact on human health.Getting understanding of the mechanisms influencing inter-microbial interactions within the microbial network,will give us an insight into the functions and properties of community as a whole.The key issues addressed by this report are how to indicate the possible interactions between species from various similar samples and how to identify species that play a key role in their community.Due to lack of standard microbial ecological network data,we used the Barabasi–Albert(BA)model,Generalized Lotka–Volterra(GLV)model and subsampling method to generate simulated microbial communities and their corresponding cross-section species-abundance tables.And we used the Leave-one method and Bray-Curtis(BC)distance formula to generate standard species keystoneness ranking according to known parameters in the simulated community.We also studied and compared network construction as well as keystoneness assessment methods based on these simulation data.Finally,we applied that method maintaining the highest degree of accuracy to the real data to analyze the actual microbial communities.In the network construction part,Pearson correlation,Spearman correlation and Mutual Information were combined to analyze the correlation between species.Then we used the deconvolutional neutral network algorithm to mitigate the indirect correlation caused by the transitivity of the interactions.Next,we applied Molecular Ecological Network Analyses(MENA)algorithm to select an appropriate threshold for determining the ecosystem's network structure.Finally,we proposed a new method named Quotient Value(QV)for calculating the direction and weights of interactions,which not only analyses correlations within the species-abundance table but also mines additional information that provides high accuracy in simulated data.In the keystoneness assessment section,we proposed a new algorithm named Spread Intensity(SI),which combined the analysis of topology of the ecological network and the characteristics of the microbial community to assess the keystoneness of species in it.Then we conducted a comparison of SI algorithm with the traditional methods in terms of Spearman correlation,precision rate,and minimum percentage with the standard ranking and the results showed that the SI method had obvious advantages in all aspects.At last,to prove that the method implemented in this system works well with realworld data,we analyzed the intestinal microbiological data of obese mice: First we constructed networks of the data and then we analyzed the keystone species related to obesity.Finally,we found that the results obtained by our system were similar to the results of real biological experiments,proving applicability of the system to the realworld data.