Multiple　myeloma　gene　expression　data　was　analyzed　and　Self　Organization　Prediction　Model　(SOPM)　based　on　Self-Organization　Mapping　(SOM)　networks　was　established　for　predicting　multiple　myeloma.　Effects　of　7129　genes　on　multiple　myeloma　were　analyzed　using　correlation　analysis　method　and　25　key　genes　from　7129　genes　were　identified　by　self-learning　procedure　of　SOM　networks.　25　genes　expression　data　were　applied　by　SOPM　to　classify　samples　and　to　predict　new　cases.　Results　indicated　that　SOPM　can　be　expected　to　learn　complex　rules　involved　in　gene　regulation,　find　key　genes　and　exactly　predict　about　98　percent　of　105　samples　based　on　the　knowledge　extracted　from　gene　expression　data.