Potential benefits of using online social network data for clinical studies on depression are tremendous. In this paper, we present a preliminary result on building a research framework that utilizes real-time moods of users portrayed in the Twitter social network and explore the use of language in describing depressive moods. First, we analyzed a random sample of tweets posted by the general Twitter population during a two month period to explore how depression is discussed in Twitter. We found remarkable activities related to depression in Twitter which included detailed information about Twitter users' depressed feelings, information sharing, attitudes towards depression, as well as their treatment histories. Going forward, we conducted a study on 69 participants to determine whether the use of sentiment words of depressed users differed from a typical user. We found that the use of words related to negative emotion and anger significantly increased among Twitter users with major depressive symptoms compared to those otherwise. However, no difference was found in the use of words related to positive emotion between the two groups. Our work provides several evidences that online social networks provide meaningful data for capturing depressive moods of users.