Computational roles assigned to SST neurons across different model architectures

Different modeling frameworks assign fundamentally different computational roles to SST neurons — from gain normalization in SSN models to prediction carriers in predictive coding models to credit assignment in learning models. This comparison reveals how theoretical commitments shape the assigned function of SST neurons.

{
  "papers": [
    {
      "doi": "10.1016/j.neuron.2014.12.026",
      "value": "Divisive normalization / surround suppression",
      "method": "SSN analysis",
      "metric": "SST computational role",
      "cite_key": "Rubin2015",
      "condition": "visual cortex",
      "study_system": "rate model",
      "value_source_sentence": "The stabilized supralinear network: a unifying circuit motif underlying multi-input integration in sensory cortex."
    },
    {
      "doi": "10.7554/elife.57541",
      "value": "Prediction signal carrier via dendritic inhibition",
      "method": "canonical circuit with plasticity",
      "metric": "SST computational role",
      "cite_key": "Hertag2020",
      "condition": "predictive coding",
      "study_system": "mean-field model",
      "value_source_sentence": "Learning prediction error neurons in a canonical interneuron circuit."
    },
    {
      "doi": "10.7554/eLife.22901",
      "value": "Credit assignment via apical dendrite error segregation",
      "method": "segregated dendrites",
      "metric": "SST computational role",
      "cite_key": "Guerguiev2017",
      "condition": "supervised learning",
      "study_system": "multi-compartment model",
      "value_source_sentence": "Towards deep learning with segregated dendrites."
    },
    {
      "doi": "10.1371/journal.pcbi.1006999",
      "value": "Somato-dendritic inhibition redistribution with VIP/PV",
      "method": "E-PV-SST-VIP analysis",
      "metric": "SST computational role",
      "cite_key": "Hertag2019",
      "condition": "gain modulation",
      "study_system": "mean-field model",
      "value_source_sentence": "Amplifying the redistribution of somato-dendritic inhibition by the interplay of three interneuron types."
    },
    {
      "doi": "10.1101/2024.10.06.616839",
      "value": "Feedback inhibition in winner-take-all computation",
      "method": "sWTA motif",
      "metric": "SST computational role",
      "cite_key": "Iqbal2024",
      "condition": "competitive computation",
      "study_system": "biophysical model",
      "value_source_sentence": "We observed a strong correspondence between the biophysical model and the TN hardware parameters, particularly in the roles of four key inhibitory neuron classes: Parvalbumin (feedforward inhibition), Somatostatin (feedback inhibition), VIP (disinhibition), and LAMP5 (gain normalization)."
    },
    {
      "doi": "10.7554/elife.82426",
      "value": "Working memory gate via VIP-SST disinhibition",
      "method": "disinhibition circuit",
      "metric": "SST computational role",
      "cite_key": "Shen2023",
      "condition": "working memory",
      "study_system": "rate model",
      "value_source_sentence": "Flexible control of representational dynamics in a disinhibition-based model."
    }
  ],
  "comparison_id": "model-types-sst-roles",
  "comparison_name": "Computational roles assigned to SST neurons across different model architectures",
  "comparison_type": "cross-study conflict",
  "what_it_reveals": "Different modeling frameworks assign fundamentally different computational roles to SST neurons — from gain normalization in SSN models to prediction carriers in predictive coding models to credit assignment in learning models. This comparison reveals how theoretical commitments shape the assigned function of SST neurons.",
  "homogeneity_check": {
    "caveats": "Different model architectures (rate, spiking, mean-field), different brain regions modeled, different computational objectives. This comparison reveals theoretical diversity rather than quantitative disagreement.",
    "comparable": false
  },
  "suggested_plot_type": "grouped bar"
}

[
  "paper:paper-2519d4c64162",
  "paper:paper-3a51e8e844f7",
  "paper:paper-485852ecd519",
  "paper:paper-abc07ad48b2f",
  "paper:paper-d557fa86de56",
  "paper:paper-e15053dfa039"
]

{
  "mode": "public_source_pointer_with_short_context",
  "notes": [
    "Local review repositories are read-only inputs.",
    "SciDEX stores paper metadata, structured evidence, file pointers, and short citation contexts; it does not copy full review prose."
  ],
  "source_commit_sha": "89b7e9787cd90e942b0adb531d549af3ddad30f1",
  "source_repository_url": "https://github.com/AllenNeuralDynamics/ComputationalReviewSST"
}