ICGenealogy: Mapping the Function of Neuronal Ion Channels in Model and Experiment

William F. Podlaski, Alexander Seeholzer, Lukas N. Groschner, Gero Miesenböck, Rajnish Ranjan, Tim P. VogelsbioRxiv 2016 · DOI

TL;DR

This paper introduces ICGenealogy, a framework and web interface for automatically classifying and comparing thousands of computational ion channel models. By standardizing voltage-clamp protocols and analyzing model responses, they've created a valuable resource that simplifies the search for appropriate models and facilitates the comparison of models with experimental data, paving the way for more robust and standardized computational neuroscience.

What Problem Does This Paper Address?

Computational neuroscience relies heavily on accurate models of ion channels, the fundamental components of neuronal electrical activity. However, the sheer volume of published ion channel models, coupled with a lack of standardization in their description and testing, makes it incredibly difficult for researchers to compare models, understand their relationships, or validate them against experimental data. This fragmentation hinders progress in building reliable, interpretable computational models of neural function.

Methods

The core of ICGenealogy is a standardized, automated approach to characterizing ion channel models. They developed a set of five voltage-clamp protocols designed to probe key gating characteristics (activation, inactivation, deactivation, etc.). These protocols were applied to each of the 2378 ion channel models (coded in NEURON) in their dataset. The resulting current traces were then processed using principal component analysis (PCA) to generate a "score vector" representing each model's behavior. These score vectors were then clustered using hierarchical clustering (Ward's method) to group models with similar functional properties. Additionally, they manually extracted metadata (like ion type, subtype, and ancestry) from publications to enrich the analysis.

Datasets

The primary dataset consists of 2378 voltage- and calcium-gated ion channel models coded in the NEURON simulation environment, sourced from the ModelDB database. This dataset was supplemented with manually extracted metadata from associated journal articles.

Key Insights and Ideas

  • Standardization is Key: The paper highlights how a standardized set of experimental-like protocols can reveal functional similarities and differences between models that might be obscured by varied reporting.
  • Beyond Metadata: While metadata is useful, directly comparing model behavior through standardized simulations provides a more direct and powerful way to categorize and compare models.
  • A "Genealogy" of Models: The concept of "family relations" (ancestor-descendant links) combined with functional clustering reveals how models have evolved and diverged, offering a historical perspective on model development.
  • Web Interface as a Resource: The ICGenealogy web interface is a crucial output, making this complex analysis accessible and actionable for the broader neuroscience community.

Key Results

  • They successfully categorized 2378 ion channel models into 211 distinct functional clusters based on their responses to standardized voltage-clamp protocols.
  • The ICGenealogy web interface provides an interactive platform for exploring these clusters, comparing models, and even comparing models to experimental data.
  • They demonstrated the framework's utility by analyzing a previously uncatalogued model and comparing experimental recordings from Drosophila melanogaster to existing models.
  • The study revealed that models with the same subtype label could fall into different functional clusters, underscoring the limitations of relying solely on nomenclature.

Limitations and Future Work

  • Model Fitting Variability: The authors acknowledge that a significant portion of model variability might stem from differences in model fitting and idiosyncratic implementation choices, rather than purely biological diversity.
  • Scope of Analysis: The current analysis was limited to voltage- and calcium-dependent ion channels coded in NEURON. Future work could extend this to other ion channel types and simulation environments (e.g., NeuroML).
  • Integration with Other Databases: The authors suggest that integrating with other databases (like IUPAR/BPS Guide to Pharmacology or Channelpedia) would further standardize nomenclature and enhance the resource.
  • Sophisticated Clustering: The continuous nature of the model scores suggests that more advanced machine-learning approaches could be applied for clustering.
  • User Submissions: The authors plan to incorporate user submissions to refine ancestral relations and improve the database.

Summary updated

Abstract

Summary Ion channel models are the building blocks of computational neuron models. Their biological fidelity is therefore crucial for the interpretability of simulations. However, the number of published models, and the lack of standardization, make the comparison of models with one another and with experimental data difficult. Here, we present a framework for the automated large-scale classification of ion channel models. Using annotated metadata and model responses to a set of voltage-clamp protocols, we assigned 2378 models of voltage- and calcium-gated ion channels coded in NEURON to 211 clusters. The IonChannelGenealogy web interface provides an interactive resource for the categorization of new and existing models and experimental recordings. It enables quantitative comparisons of simulated and/or measured ion channel kinetics, and facilitates field-wide standardization of experimentally-constrained modeling.
BibTeX
@misc{podlaski2016icgenealogy,
  author = {William F. Podlaski and Alexander Seeholzer and Lukas N. Groschner and Gero Miesenböck and Rajnish Ranjan and Tim P. Vogels},
  title = {{ICGenealogy: Mapping the Function of Neuronal Ion Channels in Model and Experiment}},
  year = {2016},
  doi = {10.1101/058685},
  howpublished = {bioRxiv},
}

Loading related content...

Figure