Due to involvement in changing of chromatin structure, histone modifications are very important in regulation for gene expression. Some attempts have been done to resolve the relationship among histone modifications and gene expression. In this study, four types of Bayeasian network methods, Maximum weight spanning tree (MWST), K2, Markov Chain Monte Carlo (MCMC) and greedy-search(GS), are used to study causal relationships. On a same dataset, the four methods resulted in different characteristics in causal relationships. MWST and MCMC were fast, but they did not give complete results, and MCMC were not stable. The initial networks had an influence on K2 and GS. In our test, GS could find some relationships that is supported by experiments.