| title | author | date | output |
|---|---|---|---|
Reproducible Research in Bioinformatics and Biostatistics |
David Fredman |
22 January 2016 |
html_document |
The scientific process in life science research, from data collection and processing to analyses, visualization and conclusions, increasingly depend on algorithms and computation. Collaborative projects where researchers share written analyses, data files and computing scripts electronically are now the norm rather than the exception. Frequently, the analysis is so complex that it is near impossible to fully describe in a written document.
To convey findings with clarity, making it possible for a researcher to completely follow someone else’s work, is the crux of scientific collaboration. Thus, reproducible research, and reproducible computation in particular, has received increasing attention in the scientific community in recent years.
In this practical and hands-on course, we will learn effective methods for collaborative and reproducible research, while learning bioinformatics and statistics methods relevant for our own research. Following a workshop that introduces and teaches methods, each student will reproduce a chosen paper using R + open source tools, and make the reproduction available online.
-
You will learn good practise for how to make your own research reproducible.
-
Reproduce the figures, tables, and statistical methods of a selected published paper using R markdown (and open-source tools).
-
Make the reproduction publically available (git repo + blog post, possibly a paper in http://rescience.github.io/read/ if suitable)
-
Understand and apply methods relevant for your own research
-
Collaborate and create science.
As part of the course, you will actually reproduce the main findings, tables and figures of a paper. You will spend substantial time and effort on reproducing a paper, so selection is very important. Please make suggestions here.
You should select two papers that you would like to reproduce (one is backup). Criteria:
-
highly relevant to your own research
-
focus mostly on analyses of empirical data
-
raw data must be available.
-
reproduction should appear feasible in a weeks effort (spread over a couple of months) at most. If the paper is massive, you may reproduce a suitabily sized and interesting section of it.
-
Collaborative science with Git [basics, branching, pull requests]
-
Data management, processing, cleaning
-
Jupyter Notebooks, RMarkdown
-
Dealing with dependencies [packaging, containers]
-
Presentation of the paper you have chosen: underlying data, which portions of the paper you will reproduce
The capstone project represents the majority of the course work. This will be done by way of online collaborative reproduction with 1-2 students per project via github.com. Bi-weekly peer-assessment [you provide feedback to other projects] and a final delivery deadline.
-
Reproductions will be in Jupyter notebooks or RMarkdown using python/R, and should be generously commented.
-
Data manipulations must be done programmatically using Python/R (and open source software if required). No alteration of original data files allowed.
-
A reproduction will follow the sequence data download, data processing, and finally reproduction of results/figures/tables.
The final report (blog post) will include:
-
A fully reproduced paper that can be generated from the raw datasets, available online here.
-
An online report of your reproduction, with code.
-
Research ideas.
-
Outstanding issues.
-
Parts of the paper we decided to not reproduce and why.
Anyone can read and contribute to a reproduction on github, see here for detailed instructions.
Please use the RRIB github repository Issues and Wiki for as much communication as possible. There will be no email list. Any enquiries should be made as an Issue or wiki entry here: https://github.com/davfre/RRIB
Norbis Students can gain 5 (10?) ECTS for attending the workshop, completing the capstone project, and actively providing peer assessment.