BackgroundResponse inhibition is a key neurocognitive factor contributing to impulsivity in mood disorders. Here, we explored the common and differential alterations of neural circuits associated with response inhibition in bipolar disorder (BD) and unipolar disorder (UD) and whether the oscillatory signatures can be used as early biomarkers in BD. Methods39 patients with BD, 36 patients with UD, 29 patients initially diagnosed with UD who later underwent diagnostic conversion to BD, and 36 healthy controls performed a Go/No-Go task during MEG scanning. We carried out time-frequency and connectivity analysis on MEG data. Further, we performed machine learning using oscillatory features as input to identify bipolar from unipolar depression at the early clinical stage. ResultsCompared to healthy controls, patients had reduced rIFG-to-pre-SMA connectivity and delayed activity of rIFG. Among patients, lower beta power and higher peak frequency were observed in BD patients than in UD patients. These changes enabled accurate classification between BD and UD with an accuracy of approximately 80 %. ConclusionsThe inefficiency of the prefrontal control network is a shared mechanism in mood disorders, while the abnormal activity of rIFG is more specific to BD. Neuronal responses during response inhibition could serve as a diagnostic biomarker for BD in early stage.