Date: April 21st 2016, 15:00-18:00
Location: Room A2-81.01 (3-01), Thorvaldsensvej 40, Frederiksberg, University of
Synopsis: This afternoon symposium gathered together researchers who are building and using open source tools to conduct reproducible science. We discussed the advantages and difficulties with this approach, and explored the best ways to implement it in practice.
The event was co-organized by IDA Biotechnology, the Danish Biotechnological Society (DBS) and CBioVikings - ISCB Regional Student Group Denmark. Event sponsors were intomics and Palantir.
Kai is an active contributor to many open source projects including Biopython and Samba, and is an advocate for open science approaches. His talk will discuss his personal approach to reproducible research.
Geoff Macintyre (Research Associate, Cambridge Cancer Center, University of Cambridge, UK) [pdf]
Geoff is a computational biologist who models tumor evolution. His talk will focus on concrete ways in which reproducible science benefits the careers of the researchers who implement it.
Sabarinathan Radhakrishnan (Biomedical Genomics Group, University Pompeu Fabra, Spain) [pdf]
Sabari studies the role of somatic mutations across different cancer types. In this talk he will focus on practical ways to enhance our research work by making it more transparent and reproducible.
Here is our full program.
Start time Title and Speaker
Alexander Junge (CBioVikings)
Helen Cook (CBioVikings)
15:20 Practice what you preach: reproducible research at the front lines of science
15:50 Sponsor Talks
16:40 Five selfish reasons to work reproducibly
Geoff Macintyre [pdf]
17:10 How to make our research work transparent and reproducible
Sabarinathan Radhakrishnan [pdf]
Jamshed Gill (CBioVikings)
Why is reproducible research important to bioinformaticians?
It was reported last year that 50% of biomedical studies are not reproducible, and it may be the case that the percentage is even higher. The causes for not being able to reproduce research studies are numerous, but generally relate to a lack of documentation of materials, protocols, or study design. Regardless of the cause, the impacts of studies that cannot be verified are similar: they represent wasted funds and effort, and delay translation of research to the clinic.
Reproducible analyses help to avoid these problems, since they can be easily rerun and verified by reviewers or third parties, and shared between research groups. Further, conducting research with a focus on reproducibility also gives direct benefits to the researchers themselves. For example, results can be brought more quickly up to date with less chance of human errors, which enables continuity on long term projects. Ultimately this means that higher quality science can be done faster.
Since most bioinformatics analysis is done with computers, reproducibility initially seems like an easy goal to reach. However, in practice bioinformaticians face many barriers that prevent true reproducibility, such as versioning, data management, and documentation challenges. This symposium will focus on the tools that help overcome these challenges: languages like R and Python are used to script analyses, version control systems track changes between collaborators, documentation systems like knitR and Jupyter notebooks document analyses, and container systems like Docker deploy finished projects. Young researchers will learn how to implement these tools in their own research to the benefit of their own scientific development.
(Click the images to visit their website)