The Dynamic Evolutionary Analysis And Similarity Measurement On Gene Regulatory Networks

Gene regulatory networks(GRNs)describe the interactions between genes and their products in organisms.Increasing evidence shows that many diseases,especially cancer,are relevant to the abnormal regulation of genes.Due to the complexity of gene regulation in cancer,the mechanism of their occurrence and development has not been fully clarified yet.Therefore,how to accurately prevent and diagnose such diseases remains unsolved.To solve this problem,in this thesis,we focus on the GRN's dynamic evolutionary Analysis as well as its similarity measurement from the perspective of complex network analysis.By this research,it is possible to predict the regulatory relationships between genes which would appear in the future.And then,the diseases such as cancer can be predicted via similarity measurement.This thesis' s works and innovations focus on two aspects: the dynamics analysis of GRNs and their similarity measurement.The details are as follows:(1)To overcome the drawback that current related researches on the dynamics analysis of GRNs usually focus only on static networks but ignore dynamic networks,in this thesis,a dynamic GRN link prediction algorithm based on the motif transfer probability is proposed from the perspective of motif evolutionary patterns.It is showed that by learning the evolutionary patterns of motifs from historical snapshots,this algorithm can accurately predict the topological structure of GRNs which would appear in the future.(2)To overcome the drawback that current related researches on the dynamics analysis of GRNs usually focus only on unsigned networks but ignore signed networks,in this thesis,the first work is extended and a dynamic GRN evolutionary analysis framework is proposed.This framework extends the study of the evolutionary analysis of GRNs to the field of dynamic signed networks by a sign discrimination algorithm based on latent space and transfer learning.This framework can make the prediction results more closer to the real-world GRNs.(3)To overcome the drawbacks that current related researches on complex network similarity measurement usually have high computational complexities,is this thesis,a fast similarity measurement for GRNs via genes' influence power is proposed.Compared with other related algorithms,this algorithm greatly reduces the computational complexity without sacrificing the accuracy.Therefore,it can be better applied in clinic practices based on large-scale GRNs.By computing the similarities between GRNs,this algorithm can be used in areas like diseases' early diagnosis and screening.This algorithm shows its great potential in time-sensitive clinical applications.The research in this thesis overcomes the shortcomings of traditional biological researches that require a lot of time and human resource.And it is a promoting work in the field of bioinformatics.The results of this research are expected to play an important role in areas such like clinical medicine,drug research and development,biological experiments in the future.