In this paper the notion of a chemical workflow engine (CWE) is introduced to support the enactment of scientific workflows, i.e. computationally intensive tasks involving large datasets. The execution of scientific workflows require the coordinated use of a large quantity of resources and scheduling of the activities. Furthermore, these workflows may pose various dynamic features that need a special elaboration, either due to the inherent structure of the workflow, changes in the environment or interaction. The concept of CWE introduced in this paper is focused on these issues with the aim of creating a highly abstract self-evolving coordination model for workflow execution. The model is based on the foundations of classical workflow formalisms, particularly, event-driven process chains, and realized through a completely new paradigm: the chemical computational model. The coordination is based on a nature metaphor and follows the principles of the chemical reactions. The proposed coordination model is inherently concurrent, exploits all potential parallelism, able to dynamically react to any changes in the environment or workflow structure.