ActivePapers - computational science made reproducible and publishable

ActivePapers is a research and development project whose aim is to make computational science more open and more reliable, by making computational reproducible and publishable. It is also a file format for storing computations.

For a detailed account of the ActivePapers project, see the paper published in F1000Research (free access).

The ActivePapers design criteria

An ActivePaper is a file combining datasets and programs working on these datasets in a single package, which also contains a detailed history of which data was produced when, by running which code, and on which machine. It is a complete record of the state of a computational research project that can be shared among collaborators and in the end published as supplementary material to a journal article.

The development of ActivePapers was guided by the following principles:

1. Data is more important than software

Computational scientists usually first choose the software for doing their work, and then let the software decide how to store the data they work on. But data is both more fundamental and more long-lived than software. A protein structure or a temperature field will make as much sense in fifty years as it does today, but the software used to produce it will probably be obsolete by then. Data should be stored in well documented and software-independent formats.

2. Code is data

We tend to think of code as the “active” part of computation and consider data the “passive” one: computation works on input data and generates output data. But code is really just yet another form of data: a set of instructions that define a computation. And since code is just data, why not store it along with the data it works on?

3. Workflows are code

A workflow is nothing but the outermost layer of the algorithm that defines a computational research project. It’s code, and thus data, and should be stored with the other data of the project.

4. One project, one file

The typical setup in computational science consists of a directory per project containing all the data, scattered over many files, some of the code, the rest being “installed” elsewhere on the computer. Workflow information is sometimes stored in the scientist’s brain (interactive shell commands, actions in a graphical user interfaces) rather than explicitly in the computer. For any non-trivial project, this quickly turns into a mess. Which data files are up to date? Which version of the program A was installed when script B was run? On which computer was file X produced? Who changed script C yesterday? All this confusion can be eliminated by storing all project information in a single file, with the consistency of the contents verified by the computer. This single file can be copied between machines, shared among collaborators, and in the end published for consultation and reuse by fellow scientists around the world.

5. Re-use is a form of citation

Scientists have always built on other scientists’ work. In computational science, this means re-using other scientists’ datasets and software. This works smoothly only through precise machine-readable references, such as DOIs (Digital Object Identifiers). Another advantage of machine-readable references is the integration of data and code references into bibliometry, creating the missing incentive for scientists to actually publish data and code.

6. Security matters.

Reproducible research invites scientists to download their peers’ computational code and run it. But running downloaded code without a careful inspection is risky: the code might destroy information on the computer it runs on, or transfer sensitive information to someone else. While malicious code is at this time not a major concern in the scientific community, defective code is not uncommon. ActivePapers provide a closed universe, isolated from other data on a user’s computer and from the Internet, in which defective and malicious code can do no harm.

7. Research moves at a slower pace than computing

In the natural sciences, researchers typically consult original journal publications that are up to about thirty years old, whereas they turn to textbooks and review articles for older work. Published computational science must therefore remain useable for a few decades. Computing environments evolve at a much faster pace, which is why software requires maintenance to remain useable for more than a few years. Computational science therefore requires a [stable computing platform] (

ActivePapers in practice

There are currently two implementations of the ActivePapers concept: the Python edition and the JVM edition. The Python edition is the more immediately useful one for most computational scientists, but the JVM edition is a more complete implementation of the principles listed above.

Both ActivePapers implementations use HDF5 as the underlying storage format. An ActivePaper is thus an HDF5 file. Datasets in an ActivePaper can be inspected using many generic HDF5 tools, in particular HDFView. HDF5 has the advantage of providing compact binary storage for large datasets and efficient access to them. This makes ActivePapers suitable for working with Big Data.

The best platforms for publishing ActivePapers are currently figshare and Zenodo. You can consult the list of ActivePapers available on figshare and on Zenodo.

For keeping up to date with ActivePapers development, follow @ActivePapers on Twitter.