Detecting Critical Points Of Complex Diseases By Using Hidden Markov Model

Detecting critical points of complex diseases is very important for early diagnosis of diseases.We proposed two algorithms based on Hidden Markov Model by using multiple samples and single samples respectively.The two algorithms construct two kinds of inconsistency indicators respectively,in order to mine the critical state between health period and disease outbreak.To verify the validity of the indicators,these indices are applied to simulated data sets,acute lung injury data and breast cancer data respectively.It is successful to detect the critical point and give the early warning signal before the disease deteriorates.In addition,the results of two actual data sets are analysed.Chapter 1,Introduction.This chapter mainly introduces the background of the paper.We discusses the complexity and serious harm of complex diseases,and discusses the necessity of exploring the critical point of complex diseases.Then we introduces several methods or algorithms of exploring the critical point of complex diseases in the past research.Chapter 2,Preparatory knowledge.In this chapter,we introduce several important steps of Hidden Markov Model,critical point bifurcation theory and the method of constructing individual-specific network.The above are the algorithm basis of this paper.Chapter 3,Algorithmic design.In this chapter,we present two algorithms for mining critical points of complex diseases.One is based on critical bifurcation theory,constructs inconsistency indicators based on multiple samples,The other is based on the properties of individual-specific networks,constructs inconsistency indicators based on single sample.Both indicators can send warning signal when the disease approach critical point.Chapter 4,simulation experiment.In this chapter,a simulation network with nine nodes is constructed.First,the critical point bifurcation theory and the properties of individual-specific network are verified in the data.Then,two inconsistency indicators are applied to the simulation network data.Both of them give warning signals when the system parameters ? 0,and the critical state is successfully excavated.Chapter 5,Practical application.In this chapter,we apply the two indicators to the lung injury data of mice and human breast cancer data,and get the same conclusion:mice are in critical state from the 4th to 8th hours after exposure to phosgene,while breast cancer is from state IIB to state IIIA.In order to verify the results,a leave-one method cross-test was carried out to verify the stability of the results.In addition,we analyzed the genes that expressed the most significant differences at the critical point:when combined with survival analysis,we found that ABCA10,ADAM33 and other genes had a significant impact on breast cancer;combined with functional analysis,the pathway which these genes are found is closely related to disease.