Data science and its future in large neuroscience collaborations

Abstract

Collaborative neuroscience requires systematic data management and analysis. How this is best done in practice remains unclear. Based on a survey across collaborative neuroscience projects, we document the current state of the art focusing on data integration, sharing, and researcher training. We propose best practices and list actions and policies to attain these goals.

Figure 1.. Results of the survey

(A) Title and locations of N = 21 currently active U19s (green boxes) and contact affiliation of individual projects as well as data science leads (blue boxes). International affiliations are not shown here. (B and C) Team (B) and laboratory (C) size of the U19s. A typical U19 team has ~40 scientists across ~7 labs. (D) The main mechanisms by which data are organized are Slack/emails and network transfer. Not much is standardized. (E) Most U19s are aware of and use tools like Neurodata Without Borders (NWB) but only in a limited capacity. (F) How much code and data are published? Note the bimodal data sharing histogram. Some collaborations, like the International Brain Laboratory, share all data. (G) Data are being published at all stages of processing without clear structure. (H) A significant fraction of code has never been reviewed. (I) Fraction of theory and experimental researchers that can implement a simple algorithm. The precise question was “What fraction of your U19’s experimental researchers has reasonable coding experience? (e.g., can implement a simple algorithm like bubble sort in Python or use the Slurm engine on a cluster.)” Dots indicate means and the standard error of the mean as confidence interval. (J) Current training is done but limited in scope. Workshops are often offered through the DataJoint team. (K) Most U19s do not enforce standards, and existing standards are relatively weak.