Translational research policies aim to reshape how biomedical scientists organize, conceive, and conduct science in order to accelerate healthcare improvements and medical innovations. Yet most analyses and evaluations of these initiatives focus on measuring the outputs generated in the different stages of the research process rather than observing scientists’ research practices directly. In this article, we analyze the collaboration networks formed by the biomedical scientists participating in a large translational research initiative. Based on data derived from a large-scale survey, we examine the network configurations established by biomedical scientists to advance their research in the context of the CIBER program—a Spanish flagship initiative aimed at supporting translational research. We adopt an ego-network perspective and draw on three network attributes—network diversity, tie strength, and tie content—to understand how scientists use their interpersonal connections to mobilize tangible and intangible resources and enable the translation of scientific knowledge into practical applications. Our cluster analysis identifies a range of scientist profiles: downstream-oriented scientists, upstream-oriented scientists, and brokering scientists. It shows that the scientists participating in the CIBER program deploy different types of collaborative behavior and engage in a variety of medical innovation activities. This suggests that the results achieved by a research program aimed at supporting collaborative networks will depend on the types of networks in which the participating scientists engage. Consequently, evaluations of these programs need to capture collaboration patterns, and should focus primarily on the collaborative process rather than the outputs that emerge from the collaboration.