The invention discloses a method of identifying a gene associated with stage III primary cancer or lymph node metastasis. The genes so identified include CAV1, CST3, LIMK1, MMP2, MMP15, VEGF, ETV4, MMP9, PIK3C2B, and SERPIN1. Also disclosed are methods for diagnosis, prognosis, and treatment of cancer. The invention further discloses compositions for preventing and treating diseases.