Scientific workflow (SWF) system is gradually liberating the computational scientists from burden of data-centric operations to concentration on their decision making. However, contemporary SWF systems fail to address the variables when scientists urge to deliver new outcomes through reproduction of workflow, including not only workflow representation, but also its "context" of use. Thus, current failure is mainly due to lack of representing and managing the "context". We propose a context-oriented approach to create a SWF system more adaptive to the dynamic research environment. We present the feedbacks from interviews of multi-national virtual screening scientists, and illustrate a case study in which their decision making processes are modelled by contextual graphs as a uniform representation of knowledge, of reasoning, and of contexts. Finally, we conclude and highlight that sharing SWF intellectually with its context would make SWF as a complement to paper-based publications.