Concepts　of　depression　drugs　are　a　kind　of　important　domain　knowledge　and　need　to　be　integrated　into　the　Data-Brain,　which　is　a　multi-dimension　knowledge　framework,　for　supporting　systematic　brain　informatics　studies　on　the　depression.　Though　some　open　biomedical　knowledge　sources　have　already　provided　depression　drug　ontologies,　it　is　still　necessary　to　realize　automatic　concept　recognition　of　depression　drugs　from　biomedical　literatures　because　of　the　quick　development　of　depression-related　studies　and　the　constant　appearance　of　new　depression　drugs.　However,　various　nomenclatures　and　a　large　number　of　abbreviations　make　it　difficult　to　extract　depression　drug　concepts　precisely　only　using　existing　methods.　This　paper　proposes　a　new　method　of　concept　recognition　based　on　the　domain　relevance　measure,　in　which　new　independence　assumptions　and　the　domain　bias　function　are　defined.　The　experimental　results　show　that　both　the　precision　and　recall　of　concept　recognition　can　been　improved　obviously　comparing　with　existing　methods.