Biologically detailed models are useful tools in neuroscience, and automated methods are now routinely applied to construct and validate such models based on the relevant experimental data. The open-source parameter fitting software Neuroptimus (formerly Optimizer) was developed to enable the straightforward application of advanced parameter optimization methods (such as evolutionary algorithms and swarm intelligence) to various problems in neuronal modeling. Neuroptimus includes a graphical user interface, and works on various platforms including PCs and supercomputers. Neuroptimus currently uses various built-in cost functions and those implemented by the eFEL feature extraction library to compare the behavior of the models to the (experimental) target data. However, this approach severely limits the range of neuronal behaviors that can be targeted by the optimization. On the other hand, the popular model-testing framework SciUnit allows the implementation of tests that quantitatively evaluate arbitrary model behaviors.
The aim of the current project is to extend the open-source neural parameter optimization tool Neuroptimus so that it is able to use test scores from the SciUnit framework as the cost function during optimization. Direct applications would include the construction of detailed biophysical models of hippocampal neurons using a combination of Neuroptimus and HippoUnit, an open-source neuronal test suite based on SciUnit. All of these tools are implemented in Python.
Skills and effort: The task would probably require a full-time effort during GSoC (350h), and at least intermediate coding skills.
Mentors: The project would be supervised by members of the Computational Neuroscience laboratory at the Institute of Experimental Medicine (Budapest, Hungary), including Sára Sáray (the developer of HippoUnit) and Szabolcs Káli @szabolcs_kali (head of the laboratory), with contributions from Máté Mohácsi (the current lead developer of Neuroptimus).
Tech keywords: Python