Research On Tumor Development Based On Inference Algorithm

In the process of clonal development of tumor cells,the original cells exist uniformly in the progeny of tumor cells through reproduction.When different gene changes occur in tumor cells,they become heterogeneous tumor cell populations,that is,subclones(SPs).Classification of SPs in tumor cells based on mutation information,phylogenetic tree representation of the relationship between different SPs,and analysis of the genetic evolution process are important means to study the mechanism of tumor cell heterogeneity.The current subclonal clustering method is fuzzy to deal with isolated points caused by noise in gene data or a mutation site does not conform to the inference model,which will indirectly affect the accuracy of subclonal classification.This paper studies how to accurately classify subclones and how to construct developmental trees that can better reflect evolutionary relationships.The main work is as follows:In order to improve the accuracy of subclonal classification,UPGMA and K-medoids were combined in this paper to analyze the high heterogeneity within tumor cells,which not only ensured the compact and distinct clusters between samples,but also reduced the noise caused by nucleotide sequences obtained by gene sequencing.The experimental results using clustering index to verify its validity show that the method has more advantages than EXPANDS in subclonal classification.In terms of the construction of developmental trees,this paper takes the lead in applying genetic algorithm and ant colony algorithm to the reconstruction task of tumor subclonal phylogenetic trees.The least evolution method(ME)is a commonly used method to construct developmental trees based on distance.The least square method is used to calculate the total length of branches as scores,and the heuristic search is conducted to search for the optimal tree.In this paper,genetic ant colony operator was used to optimize the population of developmental trees,and a new pheromone updating formula was used to obtain the optimal tree.The feasibility of searching phylogenetic trees by this method was verified by experiments compared with current tree building methods.