The cell populations, such as differentiated states or tissues and organs, are thought to be heterogeneous. Any expression profile based on a tissue sample will blend the true expression profiles of its constituent cells. To overcome this problem, single-cell methods have been developed for both microarrays and RNA-seq. Single-cell gene-expression profiling allows identification and characterization of different cell types and cell subtypes. Furthermore, phenotypic heterogeneity can be studied within the same cell type. Although several studies for single cell gene expression analysis have been reported, number of the observed cells is very limited. Therefore, we and other groups have developed novel strategy of single-cell transcriptome analysis for thousands of single cells. In these approaches, single cells are put into a high-density microwell plate or droplet and lysed in situ. mRNA is then captured on barcoding microbeads and reverse transcribed. In this study, we applied single cell transcriptome analysis to characterize complex heterogeneous samples in the human cancer tissue. These results showed the distinction of the cancer cell state with tumorigenicity, the EMT, stem like cell and infiltrated leukocytes. Finally, single cell transcriptome analysis is a powerful approach for characterizing and understanding cellular diversity in cancer tissues.