Objective: Diagnosis and management of bipolar disorder (BD) are limited by the absence of available laboratory tests. We aimed to combine data from different molecular levels and tissues into a composite diagnostic and state biomarker. Methods: Expression levels of 19 candidate genes in peripheral blood, plasma levels of BDNF, NT-3, IL-6 and IL-18, leukocyte counts, and urinary markers of oxidative damage to DNA and RNA were measured in 37 adult rapid-cycling patients with BD in different affective states during a 6- to 12-month period and in 40 age- and gender-matched healthy individuals in a longitudinal, repeated measures design comprising a total of 211 samples. A composite biomarker was constructed using data-driven variable selection. Results: The composite biomarker discriminated between patients with BD and healthy control individuals with an area under the receiver operating characteristic curve (AUC) of 0.83 and a sensitivity of 73% and specificity of 71% corresponding with a moderately accurate test. Discrimination between manic and depressive states had a moderate accuracy, with an AUC of 0.82 and a sensitivity of 92% and a specificity of 40%. Conclusion: Combining individual biomarkers across tissues and molecular systems could be a promising avenue for research in biomarker models in BD.