The invention relates to the field of Bipolar Disorder. In a recent study, the inventors performed the first genome-wide analysis of DNA methylation profiles among individuals with BD type I assessed for their response to long-term treatment with Li and identified differentially methylated regions (DMRs) between GR and NR, selecting only methylation differences that were not impacted by co-prescribed treatments that are -by definition- more frequent in NR (Marie-Claire et al., 2020). They then used MS-HRM to validate 3 DMRs that discriminated GR from NR in an extended sample of 70 individuals with BD-I. They also showed, in an independent sample that combining a few clinical variables with the three DMRs improved the performance of these MS-HRM-based biomarkers to discriminate groups (Marie-Claire et al., 2023). In the present study, they improved the performance of the combination of MS-HRMbased biomarkers and clinical predictors to discriminate individuals that benefited from Li treatment (GR and PaR) from those who did not (NR) in a retrospective sample. Thus, the present invention relates to a method for predicting whether a patient achieves a response with a treatment of Bipolar Disorder (BD) comprising determining, in a biological sample from the patient the epigenetic profile of at least 4 of the differentially methylated regions (DMRs) selected in the list consisting of DMR7291, DMR17107, DMR17978, DMR24332, DMR63769, DMR79888, DMR106540 and DMR81023