Computational techniques are central in many areas of neuroscience and are relatively easy to share. This paper describes why computer programs underlying scientific publications should be shared and lists simple steps for sharing. Together with ongoing efforts in data sharing, this should aid reproducibility of research.
1 February 1997 in Network: Computation in Neural Systems
Many areas of neuroscience are now critically dependent on computational tools to help understand the large volumes of data being created. Furthermore, computer models are increasingly being used to help predict and understand the function of the nervous system. Many of these computations are complex and often cannot be concisely reported in the methods section of a scientific article. In a few areas there are widely used software packages for analysis (e.g., SPM, FSL, AFNI, BrainVoyager, FreeSurfer in neuroimaging) or simulation (e.g. NEURON, NEST, Brian). However, we often write new computer programs to solve specific problems in the course of our research. Some of these programs may be relatively small scripts that help analyze all of our data, and these rarely get described in papers. As authors, how best can we maximize the chances that other scientists can reproduce our computations or reuse our methods on their data? Is our research reproducible? Our article lists practical suggestions to maximise the reproducibility of our work.
In the formation of retinotopic maps both experimental and theoretical work implicate guidance molecules and patterned neuronal activity. A common view is that molecular cues define and activity cues refine mappings. Important insights have come from studies of the retinocollicular projection in transgenic mice, in which cues have been modified either in isolation or in combination. Mostly these have generated descriptions of endpoint mappings. The dynamics of map formation remain under-explored experimentally and computationally. We have quantified changes in the ordering of the mouse retinocollicular projection with age after making local collicular injections of fluorescent microspheres. Contour analysis shows that, at birth (P0), cells from over 80% of the retina converge on a given collicular locus; this percentage falls gradually to P4 then rapidly approaches adult values by P12. Paired injections reveal how the segregation of labelled cells depends both on injection site separation and relative orientation and also age. At P0, large anterior-posterior separations failed to produce segregated label: segregation improved with a similar timecourse to convergence to reach near adult-values by P12. An implementation of a combined activity-molecular model captures these segregation dynamics. The developmental dynamics were then studied in the nAChR-β2 −/− mouse, which has altered patterns of retinal activity in the first postnatal week that results in a more diffuse adult retinal projection. Surprisingly, both measures of map refinement (convergence and segregation) remain largely constant and imprecise in the first postnatal week – only subsequently does the projection begin to refine. Substituting nAChR-β2 −/− activity patterns for wild-type patterns [5] into the model failed to capture the biology: refinement was initially faster . By reducing the relative importance of the gradients’ contribution to the model energy (to 15% of normal) we were able to mimic the delayed refinement observed in vivo. Our model therefore predicts the altered activity patterns may affect readout of guidance cues.