Although there is an urgent need for biological diagnosis of bipolar disorder (BD), there have been no established biomarkers. Gene expression analysis in lymphoblastoid cells (LCLs) would be a promising candidate for biomarkers. In this study, 17 candidate genes were measured in the LCLs of patients with BD. Using the data of the first set of samples (13 patients with bipolar I disorder and 21 controls), three genes, ANK3, RASGRP1 and POLG1, were selected by the logistic regression analysis with a stepwise method. Using the discriminant function generated by this analysis, the first sample was discriminated with the sensitivity of 76% and specificity of 85%. By applying the same function to the second sample set (18 patients with bipolar I and 37 controls), bipolar I disorder could be discriminated from the controls with the sensitivity of 44% and specificity of 81% (χ(2)=3.97, P=0.046). This study was the first to suggest a possible role of gene expression analysis of ANK3, RASGRP1 and POLG1 in the LCLs as potential biomarkers of BD.