Stephen Eglen

Computational NeuroscienceBy Stephen Eglen, University of Cambridge.

The annual Computational Neuroscience meeting was held this July in Antwerp, Belgium. This is a well-established meeting for researchers to discuss matters around computational modelling and analysis of neuronal systems. Although computational simulation and analysis is at the heart of this field, historically there has been little evidence of sharing of code. I was pleasantly surprised, however, to discover at the meeting that many leading labs now embrace open science. Below I outline my key observations based on observing and presenting at two workshops.

Workshop 1: Recent methods and analyses for neuronal population recordings

Recent technological developments mean that it is now possible to record the spiking activity of many hundreds or thousands of neurons simultaneously. This workshop described some of these recent techniques and the challenges for data analysis. Two themes of general interest emerged in my view from the first day:

  1. People are now sharing their computational methods; most speakers at the workshop already made…

Continue Reading

Sharing code and data neuroscienceBy Stephen Eglen, Software Sustainability Institute's fellow, University of Cambridge.

Scientists are increasingly dependent on computational techniques to analyse large volumes of data. These computational methods are often tailored to the particular analysis in mind, and as such are valuable research outputs. Furthermore, unlike experimental techniques, computational methods can be easily shared. However, at least in neuroscience, computational methods are not routinely shared upon publication of associated manuscripts.

To improve this situation, we have worked with the editors of Nature Neuroscience to establish a pilot code review project. Once papers have been approved in principle for publication, authors can opt-in to the code review. The code (and data) will be checked to see if independent reviewers can reproduce key findings of the paper. The details of the code review process are outlined in theeditorial, and we have written a commentary to describe good practice for sharing of code and data. For example, we suggest the minimum requirement for sharing is that sufficient code and data be provided to regenerate a key figure/table of the paper. This follows the well-established requirements for…

Continue Reading

By Stephen Eglen, Software Sustainability Institute Fellow and senior lecturer University of Cambridge.

Late last year, I ran a workshop with the International Neuroinformatics Coordinating Facility (INCF) in Cambridge. It was regarded by all attendees as a success and it was suggested that we archive some tips for organising a small workshop. Here are those tips.

1. Get help with admin

We were incredibly lucky in that all the administration for the event was taken care of by the INCF, and in particular its program officer, Mathew Abrams. Everyone's travel plans were coordinated, and everyone stayed at the same (beautiful) college. Good admin is a fundamental part of a successful event, but it takes a lot of time to do well, so take any help you can get to ensure that your admin is done well.

2. Assemble a diverse audience

Rather than have the same people talking to each other, make sure there is plenty of new blood with people offering different perspectives. Although most attendees were neuroscientists at my workshop, an archeologist provided insightful advice and comments from his discipline. We also had a representative from the Wellcome Trust, and two editors (from Nature and PLOS).

3. Let everyone say their piece

Day one of the workshop was devoted to introductions. Everyone was invited to present a 15-20 minute talk, with plenty of time for discussion. Workshop run more smoothly, and collaborations are set up more efficiently, when…

Continue Reading

By Stephen Eglen and Laurent Gatto, Software Sustainability Institute Fellows.

R is a well-established environment for statistical computing.  It is often seen as an alternative to computing environments such as matlab or python. In this post, we give our five reasons for why we chose to use R for research.

1. Plotting

R generates beautiful graphics with minimal effort. Publication-quality plots can be rendered in a wide range of vector- and raster-based formats. Recent extensions to the plotting system allow for complex visualisations to be expressed succintly. See R Graphics Gallery for example plots along with the code that generated the plots.

2. Packaging

R comes with a robust packaging system to allow developers and domain experts to easily distribute their code. Packages come complete with documentation, vignettes (see point 3 below), and data files.  Windows and Mac users can download packages in binary form, where C and Fortran code is pre-compiled.  As of January 2013, the Comprehensive R Archive Network (CRAN) contains 5088 packages. The packaging build system is rigourous to ensure that packages will work for for other users.  Within the field of Computational Biology, the…

Continue Reading

Reader in Computational Neuroscience, Department of Applied Mathematics and Theoretical Physics, University of Cambridge


I use computational methods to analyse and simulate the development of the nervous system, in particular the visual system.  Together with the international neuroinformatics coordinating facility (INCF), I am investigating ways to encourage code sharing and reuse in neuroscience.


I study the mechanisms that guide development of the central nervous system.  In particular, I build theoretical models to help understand and predict the processes that underlie the early development of the visual system.  I also generate new computational techniques to analyse the large volume of experimental data collected by studies of spontaneous activity patterns in the nervous system.  A common theme in these ongoing research projects is their collaborative nature. Close partnership with leading international neurobiologists ensures my work is grounded by sound biological data.  Ongoing research projects include:

1. High-throughput analysis of multi-electrode array recordings.  We are developing methods for the automated analysis of spontaneous neuronal activity.  We are screening the activity of developing networks from mice with genetic and pharmacological perturbations.

2. Development of standards for sharing of electrophysiological recordings.  Sharing data within neuroscience is hindered partly by
the lack of standard formats for storing data…

Continue Reading
Subscribe to Stephen Eglen