The invention relates to a method for sieving tumor metastasis inhibiting genes by utilization of an RNA interference library, which is to utilize an in vivo transplantation tumor metastasis model of immunodeficient mouse cells to better simulate generation of tumor metastasis than a cell model, and utilize the RNA interference library to make various genes of a human genome randomly reduce expression in tumor cell strains respectively, so as to greatly strengthen the potency of cells of which the tumor metastasis inhibiting genes reduce expression to form tumor metastasis. Therefore, after cells treated by the RNA interference library are injected into the in vivo of an immunodeficient mouse, the cells of which the tumor metastasis inhibiting genes reduce expression can form a lung tumor in the condition that contrast cells can not form tumor metastasis, and novel tumor metastasis inhibiting genes can be determined and sieved by separating the lung tumor and cloning the expression reduction genes in the lung tumor. The tumor metastasis inhibiting genes are sieved in the whole genome level by the method, and are used: firstly, for further researching generation, development and an action mechanism of tumor metastasis; secondly, as biological markers of tumor for guiding formulation of a tumor treatment solution under the condition of diagnosis, and evaluating the treatment effect and detecting relapse after diagnosis; and thirdly, as a target of tumor targeting treatment to achieve the aim of preventing and treating tumor metastasis is achieved by strengthening the functions of the tumor metastasis genes.