Identifying　genes　associated　with　specific　diseases　plays　an　important　role　in　the　pathological　study,　diagnosis,　and　treatment　of　diseases.　In　this　paper,　we　propose　a　new　method　to　identify　key　genes　for　any　specific　disease-the　mutual　information　gene　network　(MIGN)-structural　key　gene　(SKG).　Considering　brain　tumors　as　an　example,　we　identified　four　types　of　37　genes　that　have　varying　＂behaviors＂　in　MIGNs　of　normal　cells　and　different　grades　of　tumor　cells,　called　SKGs,　which　are　closely　related　to　brain　tumors.　Using　SKGs　and　K-means　clustering　algorithm　for　testing,　the　test　accuracy　rate　was　approximately　94.56%.　MIGN-SKG　effectively　identifies　a　subset　of　genes　that　may　be　markers　of　disease　progression　or　therapeutic　targets　for　the　disease.　The　key　innovation　of　MIGN-SKG　is　that　it　is　unrestricted　by　differentially　expressed　genes　and　it　directly　identifies　key　genes　from　the　perspective　of　changes　in　genetic　relationships　during　disease　progression.　It　can　identify　potential　key　genes　for　any　specific　disease　as　well　as　other　dynamic　biological　systems.(c)　2022　Published　by　Elsevier　B.V.