Article dans une revue: This article provides an original framework for analyzing networks of scientific collaborations in Europe at regional level. Which are the determinants of the observed clustering phenomenon? Which is the role of the territories? The aim of this article is to provide empirical evidence in order to answer to these questions. For this, we base our analysis on scientific collaborations between European regions in eight different disciplines (e.g. Medicine, Chemistry…) over the period 2001-2011. A normalized centrality measurement is proposed. We test the impact of territorial resources and field scientific characteristics on regions centrality in each of the analyzed discipline. Firstly, the analysis highlights a strong heterogeneity between disciplines, showing the need of carrying out specific investigation for each of them. Second, the results show the different roles played by local resources, according to the disciplines. Finally, the article discusses the implications of these results in terms of science and innovation policies.