Gα is a key subunit of heterotrimeric guanine-nucleotide-binding regulatory proteins, yet its conformational dynamics are not fully understood. In this study, we developed a Transformer-based graph neural network framework, Dynamic-Mixed Transformer (DMFormer), to investigate conformational dynamics of Gαo. DMFormer achieved an AUC of 0.75 on the training set, demonstrating robustness in distinguishing active and inactive states. The interpretability of the model was enhanced using integrated gradients, identifying the Switch II as a critical motif in stabilizing the active state and revealing distinct movement patterns between GTPase and α-Helix domains. Our findings suggest that the conformational rigidity of Q205L mutant in the Switch II segment leads to persistent activation. Overall, our study showcases DMFormer as an effective tool for analyzing protein conformational dynamics, offering valuable insights into activation mechanisms of Gα protein.