Unsupervised Clustering of Gene Expression Time Series with Conditional Random Fields

A key challenge of gene expression time series research is the development of efficient and reliable probabilistic models. In response, we propose an unsupervised conditional random fields (CRFs) model for gene expression time series clustering. Conditional random fields have demonstrated superior performance over generative models such as hidden Markov models (HMMs) in terms of computational efficiency on many sequence-data-based tasks. Yet their potential has not been previously explored in this field. In the proposed model, time series data are allowed to interact with each other via a voting pool scheme while clusters are progressively formed. Experiments based on both biological data and simulated data verify the suitability of our model to gene expression data analysis via the comparison with a recent work.