Training deep neural density estimators to identify mechanistic models of neural dynamics
Mechanistic modeling in neuroscience aims to explain observed phenomena in terms of underlying causes. However, determining which model parameters agree with complex and stochastic neural data presents a significant challenge. We address this challenge with a machine learning tool which uses deep ne...
Main Authors: | , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
eLife Sciences Publications Ltd
2020-09-01
|
Series: | eLife |
Subjects: | |
Online Access: | https://elifesciences.org/articles/56261 |
_version_ | 1811236160681279488 |
---|---|
author | Pedro J Gonçalves Jan-Matthis Lueckmann Michael Deistler Marcel Nonnenmacher Kaan Öcal Giacomo Bassetto Chaitanya Chintaluri William F Podlaski Sara A Haddad Tim P Vogels David S Greenberg Jakob H Macke |
author_facet | Pedro J Gonçalves Jan-Matthis Lueckmann Michael Deistler Marcel Nonnenmacher Kaan Öcal Giacomo Bassetto Chaitanya Chintaluri William F Podlaski Sara A Haddad Tim P Vogels David S Greenberg Jakob H Macke |
author_sort | Pedro J Gonçalves |
collection | DOAJ |
description | Mechanistic modeling in neuroscience aims to explain observed phenomena in terms of underlying causes. However, determining which model parameters agree with complex and stochastic neural data presents a significant challenge. We address this challenge with a machine learning tool which uses deep neural density estimators—trained using model simulations—to carry out Bayesian inference and retrieve the full space of parameters compatible with raw data or selected data features. Our method is scalable in parameters and data features and can rapidly analyze new data after initial training. We demonstrate the power and flexibility of our approach on receptive fields, ion channels, and Hodgkin–Huxley models. We also characterize the space of circuit configurations giving rise to rhythmic activity in the crustacean stomatogastric ganglion, and use these results to derive hypotheses for underlying compensation mechanisms. Our approach will help close the gap between data-driven and theory-driven models of neural dynamics. |
first_indexed | 2024-04-12T12:04:11Z |
format | Article |
id | doaj.art-fe7139bde6d64681b1cc8eb1344b5903 |
institution | Directory Open Access Journal |
issn | 2050-084X |
language | English |
last_indexed | 2024-04-12T12:04:11Z |
publishDate | 2020-09-01 |
publisher | eLife Sciences Publications Ltd |
record_format | Article |
series | eLife |
spelling | doaj.art-fe7139bde6d64681b1cc8eb1344b59032022-12-22T03:33:45ZengeLife Sciences Publications LtdeLife2050-084X2020-09-01910.7554/eLife.56261Training deep neural density estimators to identify mechanistic models of neural dynamicsPedro J Gonçalves0https://orcid.org/0000-0002-6987-4836Jan-Matthis Lueckmann1https://orcid.org/0000-0003-4320-4663Michael Deistler2https://orcid.org/0000-0002-3573-0404Marcel Nonnenmacher3Kaan Öcal4https://orcid.org/0000-0002-8528-6858Giacomo Bassetto5Chaitanya Chintaluri6https://orcid.org/0000-0003-4252-1608William F Podlaski7https://orcid.org/0000-0001-6619-7502Sara A Haddad8https://orcid.org/0000-0003-0807-0823Tim P Vogels9David S Greenberg10Jakob H Macke11https://orcid.org/0000-0001-5154-8912Computational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Max Planck Research Group Neural Systems Analysis, Center of Advanced European Studies and Research (caesar), Bonn, GermanyComputational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Max Planck Research Group Neural Systems Analysis, Center of Advanced European Studies and Research (caesar), Bonn, GermanyComputational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Machine Learning in Science, Excellence Cluster Machine Learning, Tübingen University, Tübingen, GermanyComputational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Max Planck Research Group Neural Systems Analysis, Center of Advanced European Studies and Research (caesar), Bonn, Germany; Model-Driven Machine Learning, Institute of Coastal Research, Helmholtz Centre Geesthacht, Geesthacht, GermanyMax Planck Research Group Neural Systems Analysis, Center of Advanced European Studies and Research (caesar), Bonn, Germany; Mathematical Institute, University of Bonn, Bonn, GermanyComputational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Max Planck Research Group Neural Systems Analysis, Center of Advanced European Studies and Research (caesar), Bonn, GermanyCentre for Neural Circuits and Behaviour, University of Oxford, Oxford, United Kingdom; Institute of Science and Technology Austria, Klosterneuburg, AustriaCentre for Neural Circuits and Behaviour, University of Oxford, Oxford, United KingdomMax Planck Institute for Brain Research, Frankfurt, GermanyCentre for Neural Circuits and Behaviour, University of Oxford, Oxford, United Kingdom; Institute of Science and Technology Austria, Klosterneuburg, AustriaComputational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Model-Driven Machine Learning, Institute of Coastal Research, Helmholtz Centre Geesthacht, Geesthacht, GermanyComputational Neuroengineering, Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany; Max Planck Research Group Neural Systems Analysis, Center of Advanced European Studies and Research (caesar), Bonn, Germany; Machine Learning in Science, Excellence Cluster Machine Learning, Tübingen University, Tübingen, Germany; Max Planck Institute for Intelligent Systems, Tübingen, GermanyMechanistic modeling in neuroscience aims to explain observed phenomena in terms of underlying causes. However, determining which model parameters agree with complex and stochastic neural data presents a significant challenge. We address this challenge with a machine learning tool which uses deep neural density estimators—trained using model simulations—to carry out Bayesian inference and retrieve the full space of parameters compatible with raw data or selected data features. Our method is scalable in parameters and data features and can rapidly analyze new data after initial training. We demonstrate the power and flexibility of our approach on receptive fields, ion channels, and Hodgkin–Huxley models. We also characterize the space of circuit configurations giving rise to rhythmic activity in the crustacean stomatogastric ganglion, and use these results to derive hypotheses for underlying compensation mechanisms. Our approach will help close the gap between data-driven and theory-driven models of neural dynamics.https://elifesciences.org/articles/56261bayesian inferencedeep learningstomatogastric ganglionmodel identificationneural dynamicsmechanistic models |
spellingShingle | Pedro J Gonçalves Jan-Matthis Lueckmann Michael Deistler Marcel Nonnenmacher Kaan Öcal Giacomo Bassetto Chaitanya Chintaluri William F Podlaski Sara A Haddad Tim P Vogels David S Greenberg Jakob H Macke Training deep neural density estimators to identify mechanistic models of neural dynamics eLife bayesian inference deep learning stomatogastric ganglion model identification neural dynamics mechanistic models |
title | Training deep neural density estimators to identify mechanistic models of neural dynamics |
title_full | Training deep neural density estimators to identify mechanistic models of neural dynamics |
title_fullStr | Training deep neural density estimators to identify mechanistic models of neural dynamics |
title_full_unstemmed | Training deep neural density estimators to identify mechanistic models of neural dynamics |
title_short | Training deep neural density estimators to identify mechanistic models of neural dynamics |
title_sort | training deep neural density estimators to identify mechanistic models of neural dynamics |
topic | bayesian inference deep learning stomatogastric ganglion model identification neural dynamics mechanistic models |
url | https://elifesciences.org/articles/56261 |
work_keys_str_mv | AT pedrojgoncalves trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT janmatthislueckmann trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT michaeldeistler trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT marcelnonnenmacher trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT kaanocal trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT giacomobassetto trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT chaitanyachintaluri trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT williamfpodlaski trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT saraahaddad trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT timpvogels trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT davidsgreenberg trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics AT jakobhmacke trainingdeepneuraldensityestimatorstoidentifymechanisticmodelsofneuraldynamics |