The parallel coordinate (PC) plot is a powerful visualization tools for high-dimensional data. In this paper, we explore its usage on gene expression data analysis. We found that both the additive-related and the multiplicative-related coherent genes exhibit special patterns in the PC plots. One-dimensional clustering can then be applied to detect these patterns. Besides, a split-and-merge mechanism is employed to find the biggest coherent subsets inside the gene expression matrix. Experimental results showed that our proposed algorithm is effective in detecting various types of biclusters. In addition, the biclustering results can be visualized under a 2D setting, in which objective and subjective cluster quality evaluation can be performed.