[go: up one dir, main page]
More Web Proxy on the site http://driver.im/ Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2013 Jul 31.
Published in final edited form as: Neuroinformatics. 2003;1(1):135–139. doi: 10.1385/NI:1:1:135

ModelDB: making models publicly accessible to support computational neuroscience

Michele Migliore 1,2, Thomas M Morse 1, Andrew P Davison 1, Luis Marenco 3, Gordon M Shepherd 1, Michael L Hines 1
PMCID: PMC3728921  NIHMSID: NIHMS489726  PMID: 15055399

Computational neuroscience as a scientific discipline must provide for the ready testing of published models by others in the field. Unfortunately this has rarely been fulfilled. When exact reproduction of a model simulation is achieved, it is often a long and difficult process. Too often, missing or typographically incorrect equations and parameter values have made it difficult to explore or build upon published models. Compounding this difficulty is the proliferation of platforms and operating systems that are incompatible with the author's original computing environment.

Because of these problems, most models are never subjected to the rigorous testing by others in the field that is a hallmark of the scientific method. This not only impedes validation of a model, but also prevents a deeper understanding of its inner workings, especially through modification of the parameters. Furthermore, modular pieces of the model, e.g. ion channels or the morphology of a cell, cannot be reused to build new models and propel research forward.

ModelDB (http://senselab.med.yale.edu/modeldb) is intended to address these issues (Peterson et al, 1996; Shepherd et al, 1998). ModelDB is a database of computational models, either classics in the field or published in recent years. It focuses on models for different types of neurons, and presently contains over 60 models for 15 neuron types. In addition to compartmental models, it contains models covering from ion channels and receptors through axons and dendrites through neurons to networks. Models can be accessed by author, model name, neuron type, concept, e.g. synaptic plasticity, pattern recognition, etc, or by simulation environment.

ModelDB is a member of a major neuroscience database collection called SenseLab. Each SenseLab database has an easily extensible structure achieved through the EAV/CR (Entity-Attribute-Value with Classes and Relationships) data schema (Nadkarni et al 1999, Miller et al 2001). ModelDB is integrated with NeuronDB (Marenco et al 1999), another SenseLab database that stores neuronal properties derived from the neuroscience literature (http://senselab.med.yale.edu/senselab/NeuronDB). Use of the models is free to all. Contributing to the database is also open to all. Contributions are tested for quality-control purposes before being made public. Here we describe how to find, run, and submit models to ModelDB.

Browsing ModelDB

The use of ModelDB typically starts with a computational neuroscientist who wishes to test the results of a simulation by a published model, and use that as a starting point for further research. Instead of recreating the model from scratch, the user goes to the ModelDB home page (Fig.1, top), to find the model by any of the various ways already mentioned. All the information about a model is shown on a single page (Fig.1, bottom), which also contains tools for finding related models in the database.

Figure 1.

Figure 1

Open in a new tab

(Top) The home page of ModelDB. (Bottom) Typical appearance of a model's home page. In this case, a model for thalamic relay neurons is shown.

The files required to run a simulation are stored in a compressed archive in zip format and can be browsed and/or downloaded (Fig.1, bottom, lower left column). There are links between ModelDB and NeuronDB, to show extensive information about the distribution of receptors, currents, and transmitters, derived from the experimental literature for the particular neuron or property to which the model refers. For models implemented in the NEURON simulator (Hines and Carnevale, 1997), the simulation can be run directly from the browser (by clicking on the “Auto-launch” button), if NEURON is already installed on the host computer.

Submitting Models

The value of a database depends on it being populated with data. In order to stimulate input from the community we summarize briefly the submission process. Details may be found at http://senselab.med.yale.edu/senselab/ModelDB/guide2.html.

A model submission consists of sending the reference to a paper (each model in the database must be associated with a paper published in a peer-reviewed journal) and a compressed archieve (e.g. zip file) containing the model files that are needed to reproduce some of the results reported in the paper. The administrator will check the files for consistency, confirm that the model reproduces any intended figure(s), and then add the model to the database.

Contributors decide which simulations best illustrate the model. According to the kind of model that one wants to submit, there are a few obvious choices. For example, realistic implementations of a specific neuron, e.g. the model of thalamic relay neurons (Destexhe et al, 1998, http://senselab.med.yale.edu/senselab/ModelDB/ShowModel.asp?model=279), can show the firing properties under simple current injections. More elaborate models, such as those dealing with network properties (e.g. synchronization), higher brain functions (e.g. memory or vision), or specific computational mechanisms (e.g. dendritic integration, temporal summation, or synaptic plasticity) require a careful selection of what to demonstrate. A user should be able to grasp quickly and simply the main features and capabilities of the model. Auto-launch demonstrations that run longer than one minute tend to tax users. Complex plots, drawn during or at the end of long simulations, are justified only when they are the most important part of a model. One of the latter examples is the reduced implementation of the Hopfield and Brody (2001) model http://senselab.med.yale.edu/senselab/ModelDB/mdbFtof.asp?model=2798), illustrating how a transient synchrony of many neurons in a network can recognize a spatiotemporal activity pattern.

Rather than sending files to the administrator, users may prefer to maintain complete control of the entire submission process. In this case, the ModelDB administrator creates a private account in which authors can upload and test different ways to present model(s) before making it (them) public.

Model files are often used by investigators as a starting point to arrange their own models and simulations. To encourage the dissemination and use of the models, contributors should consider simplifying, modularizing, and documenting their code as much as possible. Adding a new model and linking it to the SenseLab searching facilities is straightforward. A quick look at several public models will suggest how to define the various fields. Every aspect of the model (name, notes, properties, and the archive file via re-uploading) can be edited as many times as necessary, until the contributor decides that the model is now ready to be made public. Even after a model is made public it can still be edited.

ModelDB has played an essential role in research on the active properties of dendrites (Shen et al, 1999; Migliore and Shepherd, 2002). Other investigators have used it for creating a network output for driving a particular cell (Chemin et al, 2001), and for tutorials for developing computational modeling skills. Worldwide, many institutions have downloaded files from ModelDB.

Conclusion

ModelDB provides a resource for the computational neuroscience community that enables investigators to increase their understanding of published models by enabling them to run the models as published and build on them for further research. Its use can aid the field of computational neuroscience to enter a new era of expedited numerical experimentation.

Acknowledgments

We would like to thank the community of ModelDB users for their patience and suggestions. This work has been supported by the National Institute on Deafness and Other Communication Disorders, National Institute of Mental Health, National Institute of Neurological Disorders and Stroke, National Institute on Aging, and National Science Foundation (Human Brain Project), and a Multi University Research Initiative (Department of Defense).

References

  1. Chemin J, Monteil A, Bourinet E, Nargeot J, Lory P. Alternatively spliced alpha(1G) (Ca(V)3.1) intracellular loops promote specific T-type Ca(2+) channel gating properties. Biophys J. 2001;80:1238–50. doi: 10.1016/S0006-3495(01)76100-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Destexhe A, Neubig M, Ulrich D, Huguenard J. Dendritic low-threshold calcium currents in thalamic relay cells. J Neurosci. 1998;18:3574–8358. doi: 10.1523/JNEUROSCI.18-10-03574.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Hines ML, Carnevale NT. The NEURON simulation environment. Neural Comput. 1997;9:1179–1209. doi: 10.1162/neco.1997.9.6.1179. [DOI] [PubMed] [Google Scholar]
  4. Hopfield JJ, Brody CD. What is a moment? Transient synchrony as a collective mechanism for spatiotemporal integration. Proc Natl Acad Sci USA. 2001;98:1282–1287. doi: 10.1073/pnas.031567098. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Marenco L, Nadkarni P, Skoufos E, Shepherd G, Miller P. Neuronal database integration: the Senselab EAV data model. Proc AMIA Symp. 1999:102–106. [PMC free article] [PubMed] [Google Scholar]
  6. Migliore M, Shepherd GM. Emerging rules for the distributions of active dendritic conductances. Nat Rev Neurosci. 2002;3:362–70. doi: 10.1038/nrn810. [DOI] [PubMed] [Google Scholar]
  7. Miller PL, Nadkarni P, Singer M, Marenco L, Hines M, Shepherd G. Integration of multidisciplinary sensory data: a pilot model of the human brain project approach. J Am Med Inform Assoc. 2001;8:34–48. doi: 10.1136/jamia.2001.0080034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Nadkarni PM, Marenco L, Chen R, Skoufos E, Shepherd G, Miller P. Organization of heterogeneous scientific data using the EAV/CR representation. J Am Med Inform Assoc. 1999;6:478–93. doi: 10.1136/jamia.1999.0060478. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Peterson BE, Healy MD, Nadkarni PM, Miller PL, Shepherd GM. ModelDB: an environment for running and storing computational models and their results applied to neuroscience. J Am Med Inform Assoc. 1996;3:389–98. doi: 10.1136/jamia.1996.97084512. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Shen GY, Chen WR, Midtgaard J, Shepherd GM, Hines ML. Computational analysis of action potential initiation in mitral cell soma and dendrites based on dual patch recordings. J Neurophysiol. 1999;82:3006–20. doi: 10.1152/jn.1999.82.6.3006. [DOI] [PubMed] [Google Scholar]
  11. Shepherd GM, Mirsky JS, Healy MD, Singer MS, Skoufos E, Hines MS, Nadkarni PM, Miller PL. The Human Brain Project: Neuroinformatics tools for integrating, searching, and modeling multidisciplinary neuroscience data. Trends Neurosci. 1998;21:460–468. doi: 10.1016/s0166-2236(98)01300-9. [DOI] [PubMed] [Google Scholar]

RESOURCES