Is realistic neuronal modeling realistic?

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Biophysics

Background:

• Scientific models simplify reality to explain phenomena and predict experiments.
• Recent advances in computational neuroscience focus on detailed, realistic modeling of single neurons.
• The discovery of complex neuronal computations beyond simple input summation drives this field.

Purpose of the Study:

• To critically evaluate the feasibility and philosophical implications of realistic single neuron modeling.
• To assess the performance of current compartmental models of layer 5 neocortical pyramidal neurons.
• To identify limitations and propose future directions for developing more accurate neuronal models.

Main Methods:

• Review of the state-of-the-art in compartmental modeling, including progress and limitations.
• Testing three publicly available models of layer 5 pyramidal neurons with computational challenges.
• Analysis of model performance and comparison with published data.

Main Results:

• Current compartmental models are often ad hoc and perform poorly outside their designed applications.
• These models exhibit unrealistic behavior when stretched beyond specific computational problems.
• A survey of published models suggests widespread issues with generalizability and realism.

Conclusions:

• Practically, current realistic single neuron models are limited in their accuracy and scope.
• Philosophically, the pursuit of hyper-realistic models may not always yield better understanding.
• New strategies are needed to develop more robust and genuinely realistic single neuron models.