Distinguishing cancer cells from normal cells is important in both biological research and clinical applications. Various techniques have already been developed to achieve single-cell classifications such as using target anti-body for cell type classification. However, there are still limitations adopting biological methods such as low-throughput and high cost. Physical properties of cells has been reported to be different between cancerous cells and normal cells. Taking physical properties as indicators for cancer cell classification is proved to be effective and inexpensive. It is known that certain proteins such as actin and lamins in cytoskeleton and nucleoskeleton determine cell elasticities and proteins such as integrin plays important roles in cell-extracellular matrix (ECM) adhesion forces. Therefore, regarding elasticities of cytoskeleton and cell-ECM adhesion forces can be a promising approach to realize effective cancer cell classification. Herein, three methods were investigated for distinguishing cancer cells: (1) an elasticity modulatable calcium-alginate microbead was developed as an elasticity indicator to distinguish floating cancer cells; (2) a non-destructive nuclear elasticity measurement was established for high-throughput nuclear stiffness examination and cell classification; (3) investigating cell initial spreading behavior on different ECM proteins coating to differentiating cancer cells. To extend application of synthetic microbead, cell classifications were not only conducted in culture medium environment but also in diluted whole blood environment. Success of this classification scheme takes advantages of whole cell elasticity differences between cancer cells and normal cells, which may mainly be contributed by both cytoskeleton and nucleoskeleton. On the other hand, a higher significant deformability difference was found between nasopharyngeal cancerous cells and normal cells. Thus, a direct technique was developed to quantify nuclear elasticity of nuclei in floating cells. Importantly, a simple physical model was built for whole cell bodies to realize non-destructive quantification of nuclear stiffness. Considering a significant measuring difference between nuclear elasticity in intact cells and isolated environment, staining of Lamin A/C was conducted to investigate why elasticity difference was exist, which convince the non-destructive method could obtain more accurate nuclear elasticity. Although these two methods were proved to be feasible for cell classification, distinguishing cancer cells at adherent condition is yet to be achieved. As initial cell adhesion process is reported to be much determined by cell physical properties, observing initial cell spreading behavior can be a simple way for cancer cells distinguishing. Methods reported in this thesis could provide better classifications of cancer cells in single-cell level and give better understandings of how physical properties interact with cell behaviors.