Identifying the pre-transition state during biological processes by hidden Markov model

Identifying the pre-transition state just before the occurrence of a critical transition during a complex biological process is a challenging task, because the state of the system may show little apparent change or clear phenomenon before this critical transition during the biological processes. By regarding that the pre-transition state is the end or change-point of a stationary Markov process, we present a novel computational method, hidden Markov model (HMM) based state-transition forward method, which is a non-parametric estimation and can identify the pre-transition state. To validate the effectiveness, we apply this method to detect the signal of the imminent critical deterioration of complex diseases based on both simulated dataset and the rich information provided by high-throughput microarray data. We identify the pre-transition states and a number of related modules for the acute lung injury triggered by phosgene inhalation. Both functional and pathway enrichment analyses validate the results.