A　major　focus　of　cancer　research　is　to　infer　informative　cancer　gene　association　networks　from　gene　expression　data.　We　introduced　the　relevance　network　to　find　the　cancer　genes＇　association　that　may　represent　prognostic　factors　and　potential　targets　for　anticancer　therapies.　On　the　base　of　relevance　networks　using　mutual　information,　interactions　of　the　informative　genes　are　shown　graphically　and　functional　genes　are　clustered.　We　used　a　public　leukemia　data　set　of　72　RNA　expression　samples　of　50　genes　to　construct　relevance　networks.　Several　relevance　networks　were　produced.　The　biological　significance　of　relevance　networks　Is　explained.　These　interactions　between　the　genes　reveal　the　mechanism　of　leukemia　and　the　correlated　genes.　The　results　show　that　the　method　can　be　used　to　find　functional　genomic　clusters　and　inferring　cancer　genes＇　association　networks,　independent　of　previous　biological　knowledge.