From Bio-jETI Process Models to Native Code

Bio-jETI is a framework for model-based, graphical development and execution of bioinformatics analysis processes. With the GeneSys code generation framework we can automatically compile the workflow models into native, stand-alone program code. We show via a phylogenetic analysis workflow designed by the DNA Data Bank of Japan (DDBJ) how we generate 6 variants of Java code from the corresponding process model realized in Bio-jETI. Performance measurements show that 1) the overall workflow execution time is dominated by the remote services it uses, and thus 2)all 6 variants are almost as fast as the handwritten Java of DDBJ. This way, we obtain efficient native code essentially without programming. Thus, we demonstrate in this paper that model- based workflow development in Bio-jETI offers several advantages over manual implementation - including higher agility, greater transparency and better maintainability - without compromising the runtime performance.