Learning Hybrid Biophysical Neuron Models with Neural ODEs
Researchers have developed a novel hybrid modeling framework that integrates neural ordinary differential equations (Neural ODEs) into biophysical neuron models. This approach allows for the flexible discovery of unknown or mis-specified ion channel kinetics directly from voltage recordings, preserving mechanistic interpretability. The method has demonstrated its ability to fit existing ion channel models and recover unknown gating dynamics, even generalizing to different stimulus regimes. Furthermore, it can reduce the computational cost of complex neuron models, such as a multicompartment cortical neuron model, by creating a single-compartment hybrid model with learned axial current. AI
IMPACT This research introduces a novel method for enhancing biophysical neuron models, potentially accelerating neuroscience research by improving model accuracy and efficiency.
- Hugging Face
- arXiv
- Quantitative Biology
- Neural ODEs
- Neural Ordinary Differential Equations
- Neurons and Cognition
- Biophysical neuron models
- ion channel complex
- voltage recordings
- current-clamp recordings
- Multicompartment model of body composition assessment in Chinese-American adults
- cortical neuron
- single-compartment hybrid model
- axial current