Research On Path Extraction Algorithm In Mining Microbial Interaction Relationship

In nature,most microbes exist in complex and diverse community ecosystems that interact and intertwin to maintain the biological functions and ecological roles of micro-ecosystems.How to accurately and reliably discover important microbial interactions from these community ecosystems is a hot topic in recent years.In addition,it is a new and important direction to find high-order microbial interactions and study their effects on ecosystem functions.This thesis integrates multi-omics data such as microbial abundance data,phylogenetic tree and its metabolic network data obtained by metagenomic sequencing technology,and develops a path extraction based strategy to identify and predict the interaction between microbial species and find high-order interactions.The main contributions are as follows:First,the microbial interaction prediction method based on the iterative random forest(iRF)algorithm was developed,and the microbial phylogenetic distance information was integrated to make the results more biologically significant.This method puts the microbial interaction extraction into a supervised learning framework,which ensures the reliability of the relationship extraction process and the stability of the results.Integrating microbial phylogenetic distance data ensures that the discovered interactions have real biological significance.Finally,the microbial metabolic network model was used to verify the interaction relationship from the metabolic point of view,which provided a candidate set for the biological experiment design of microbial co-culture.Secondly,a genetic high-order relationship prediction algorithm based on genetic ant colony optimization(GACO)algorithm is proposed.Using a heuristic search strategy,not only can you find microbial interactions with linear or logical additivity,but you can also infer higher-order interactions with complex connections.This paper classifies and encodes dat,utilizes distributed search strategy,and based on positive feedback mechanism,it continuously cross-compiles microbial species combinations and finally mines different types of microbial interactions.Experiments show that heuristic search strategy can effectively improve the ability to discover community microbial interactions and avoid the logical defects of traditional methods.In addition,the results also demonstrate the sparsity of high-order interactions in the microbial community ecosystem..In this thesis,based on the path extraction strategy,two methods of discovering or inferring microbial interactions are studied,and proposes a complete process for integrating multi-omics microbial data,from interaction relationship extraction to result verification,,which is provided important ideas of development better calculation methods and design biological experiments for identification of microbial interactions in the future.