Google Meet link: https://meet.google.com/nnj-tosy-fec
Sára Sáray,
Andrew Davison,
Szabolcs Kali,
Christian Rössert,
Shailesh Appukuttan,
Eilif Muller,
Tamás FreundAnatomically and biophysically detailed neuronal models, that are built in a data-driven manner, are useful tools in understanding and predicting the behavior and function of the different cell types of the brain. Due to the growing number of computational and software tools and the increasing body of experimental data from electrophysiological measurements, that enable more accurate neuronal modeling, there is a constantly increasing number of different models of many cell types available in the literature. These are usually developed using different methods and for different purposes, most often to reproduce the results of a few selected experiments, and it is often unknown how they would behave in other situation or whether they are able to generalize outside their original scope. This might be the reason why it is uncommon in the modelling community to re-use and further develop already existing models, which prevents the construction of consensus “community models” that could capture an increasing proportion of the electrophysiological properties of the given cell type. In addition, even when models are re-used they may lose their ability to capture their originally adjusted behavior while their parameters are retuned to make them fit another subset of experimental data.
The collaborative approach of model development requires extensive validation test suites which enables modelers to evaluate their models against experimental observations according to standardized criteria and to explore the changes in model behavior at the different stages of its development. Applying automated tests also facilitates optimal model re-use and co- operative model development by making it possible to learn more about models published by other groups (beyond the results included in the papers) with relatively little effort.
Initially we addressed this issue by developing an open-source Python test suite, called HippoUnit () for the automated and systematic validation and quantitative comparison of the behavior of models of the hippocampal CA1 pyramidal cells (which is one of the most studied cell type) against electrophysiological data. We applied HippoUnit to test and compare the behavior of several different hippocampal CA1 pyramidal cell models available on ModelDB (results are available at: ). We also employed the test suite to aid the development of models within the Human Brain Project (HBP) and integrated the tests into the validation framework developed in the HBP.
Currently we are extending this test suite by adding new tests for the validation of other important hippocampal cell types. New validation tests cover somatic behavior and signal propagation in dendrites of basket cells and CA3 pyramidal cells, and the propagation of action potential in the axon of basket cells.
By presenting these results we hope to encourage the modeling community to use more systematic testing during model development, in order to create neural models that generalize better, and make the process of model building more reproducible and transparent.
Acknowledgments: Supported by the ÚNKP-19-3-III New National Excellence Program of the Ministry for Innovation and Technology; European Social Fund (EFOP-3.6.3-VEKOP-16-2017-00002); the European Union’s Horizon 2020 Framework Programme for Research and Innovation (Specific Grant Agreements No. 720270, 785907 - Human Brain Project SGA1, SGA2).
