This paper reports our ongoing work investigating the structural features of scientific collaboration based on metadata collected from a scientific data repository (SDR). The background literature is reviewed in supporting our claim that metadata collected from SDRs offer a complimentary data source to traditional publication metadata collected from digital libraries. Methodological considerations are discussed in association with using metadata from SDRs, including author name disambiguation and data parsing. Initial findings show that the network has some unique macro-level structural features while also in agreement with existing networks theories. Challenges due to inconsistent metadata quality control procedures are also discussed in an attempt to reinforce claims that metadata should be designed to support both domain specific retrieval and evaluation and assessment needs.