With the advancement of VR technology, meetings increasingly take place in virtual environments. Mechanisms for supporting VR-based meetings may fail without the understanding of explicit and implicit roles participants play in such meetings. The leader role in particular has a large impact on the success and failure of meetings. Since existing research shows the effectiveness of nonverbal features in role detection, we collect audio data from the members of VR-based group discussions and extract their nonverbal features so as to achieve leader detection. Moreover, we analyze the data in depth and discuss the differences and similarities in speaking activities of group leaders in VR and face-to-face settings.