Distributed collaboration over the Internet has become increasingly common in recent years, supported by various technologies such as virtual workspace systems. Often such collaboration is ad-hoc, and virtual workspaces are set up anew for each new instance of collaboration. We propose that much of the ad-hoc collaboration can be captured and transformed into patterns for reuse in future collaboration. This paper presents the results of the past five years of our work in this area. We introduce the notion of patterns of virtual collaboration; present a framework for extracting patterns of work in virtual workspace systems; and introduce an information model of virtual collaboration. We then present an overview of our data and process mining methods and reverse engineering techniques for discerning work processes carried out through virtual workspace systems. Finally we present our visual mining techniques that we use to discern aspects of work processes in virtual workspaces.