Recent educational policies advocate a radical revision of science curricula and pedagogy, to support interdisciplinary practices, a distinguishing feature of contemporary science. Computational modeling (CM) is a core methodology of interdisciplinary science, as such models allow intertwining of data and theoretical perspectives from multiple domains, to address complex problems such as climate change and pandemics. This integrative nature of CM could support the pedagogical transition to interdisciplinary science as well. Most approaches to introduce CM in science curricula are based on learning new practices, such as VPython programming or agent-based modeling. These approaches do not integrate CM with existing content, media, and teaching practices. To facilitate this integration, we present a more gradualist design, starting from derivation models in physics. This design was implemented as a set of teacher professional development modules, and presented to a group of physics teachers interested in introducing CM to undergraduate students. The analysis of their responses indicates that even this gradual transition to CM requires teachers to significantly revise their ideas about the nature of physics and physics learning (their personal epistemologies). We discuss how the teacher professional development modules were redesigned based on this finding.