Building a local community of practice in scientific programming for life scientists

For most experimental biologists, handling the avalanche of data generated is similar to self-learn how to drive. Although that might be doable, it is preferable and safer to learn good practices. One way to achieve this is to build local communities of practice by bringing together scientists that perform code-intensive research to spread know-how and good practices. Here, we indicate important challenges and issues that stand in the way of establishing these local communities of practice. For a given researcher working for an academic institution, their capacity to conduct data-intensive research will be arbitrarily relying on the presence of well-trained bioinformaticians in their neighborhood. In this paper, we propose a model to build a local community of practice for scientific programmers. First, Software/Data Carpentry (SWC) programming workshops designed for researchers new to computational biology can be organized. However, while they provide an immediate solution for learning, more regular long-term assistance is also needed. Researchers need persisting, local support to continue learning and to solve programming issues that hamper their research progress. The solution we describe here is to implement a study group where researchers can meet-up and help each other in a "safe-learning atmosphere". Based on our experience, we describe two examples of building local communities of practice: one in the Netherlands at the Amsterdam Science Park and one in the United States at the University of Wisconsin-Madison. The current challenge is to make these local communities self-sustainable despite the high turnover of researchers at any institution and the lack of academic reward (e.g. publication). Here, we present some lessons learned from our experience. We believe that our local communities of practice will prove useful for other scientists that want to set up similar structures of researchers involved in scientific programming and data science.