Provenance and reproducibility in the automation of a standard computational neuroscience pipeline

Abstract

Rapid increase in data volume, compounded by the reproducibility crisis, has led to the need to automate both experimental and computational aspects of neuroscience investigations. Automating neuroscience investigations enables an unprecedented ability to record and inspect how results were achieved. Here we review some of our recent work to integrate provenance and reproducibility measures into a tool called NeuroManager that automates a standard computational neuroscience pipeline, unifying the experiment-data-modeling-analysis cycle and allowing the scientist to focus on model evolution. Through a flexible daily workflow that leverages servers, clusters, and clouds simultaneously, NeuroManager automates manual tasks including database access, job submission, simulation scheduling, and preservation of provenance.