Much of the computational research in biology and medicine are performed on the UNIX command line, using a combination of larger programs, smaller scripts and interactive trial and error. Due to a high degree of trial-and-error, as well as high time pressure to get new findings published, computational analyses performed on the UNIX command line are often poorly documented and very difficult to reproduce (or re-use for new purposes) by independent researchers. There has lately been a renewed focus on how results achieved by one research group can be readily reproduced by other groups. This has for instance been emphasized with respect to drug discovery, where it is essential that promising results can be critically examined and reproduced by other groups before expensive clinical trials are initiated. One aspect of improving the reproducibility of computaitonal analyses lies in getting researchers to learn good habits and prioritize putting effort into reproducibility of their work. Another aspect lies in developing smart technical solutions and new processes of doing computaional analyses, so that acceptable reproducibility can be achieved with as little effort as possible
The goal of the task is to study how researchers in biology and medicine perform computational research on the UNIX command line and to suggest improvements to how this is done. Studying this process will consist of both trying to reproduce research findings of international groups based on written documentation, and studying local researchers while they are working on real research cases. Suggested improvements may be in the form of useful habits, technologies and processes that ensure reproducibility, or newly developed pieces of code that researchers may apply to increase the reproducibility of their work with little further effort.