ICG User's Manual
This is a simple instruction manual for the IonChannelGenealogy Browser, located at http://icg.neurotheory.ox.ac.uk/viewer. Please refer to the corresponding publication for more background information. This manual was last updated on November 28, 2016. Please submit a ticket http://icg.neurotheory.ox.ac.uk/contribute if you notice any errors or would like to suggest any additions.
This document is also available for download in PDF format.
Table of contents
- 1: Overall layout
- 2: Selecting a model
- 3: Type Selection
- 4: Family Selection
- 5: Color Selection
- 6: Filter Selection
- 7: Compare Models
- 8: Examples
1: Overall layout
The ICG Browser is composed of a main visualization space (Figure 1, A) and a toolbar on the right (Figure 1, B). The visualization space can be manipulated by zooming in and out using a standard mouse scroll or trackpad scroll, as well as
drag+click. A help button in the top left corner of the visualization space provides some basic information about the browser (Figure 2). Note that the right toolbar scrolls downwards to display more information. The items found in the right toolbar are highlighted in Figure 3, which will each be described in turn:
Type Selection: select between four different data viewers: hierarchy view, XY view, similarity view, and cluster view. Additional options for the current view, found below the view option box, can be turned on and off with the check box. See Section 3 for more details.
Family Selection: select among five ion type families: voltage-dependent potassium (K), voltage-dependent sodium (Na), voltage-dependent calcium (Ca), calcium-dependent potassium (KCa), and voltage-dependent afterhyperpolarization-activated cation (Ih). All four data views in type selection can be used for each family. Note that only the K family is shown in this instruction manual. See Section 4 for more details.
Color Selection: select coloring of nodes by a metadata item. See Section 5 for more details.
Filter Selection: filter the currently visible nodes by a particular metadata item. See Section 6.
Channel list: list of all channel models currently being displayed, with id, mod name, and modelDB id. Scroll down to view. Click on a model to highlight node in visualization space and to go to channel details. See Section 2 for more details.
2: Selecting a model
Models can be selected by clicking on individual items in the visualization space in any of the three views. Hovering over a model brings up a tooltip box (Figure 4, A) which contains a small summary of the model, including model ID, model name, publication reference, modelDB ID, subtype, cluster membership and a small image of the current response traces.
Clicking on a model brings up the Channel Details on the right (Figure 4, B) in place of the channel list. The channel details lists all metadata information and is described below:
- ID: unique identification number for each channel model on ICGenealogy
- Name: file name of the mod file, as taken from ModelDB entry
- ModelDB ID: Identification number associated with each entry on ModelDB. All channel models from the same entry have the same ID number.
- Reference Information: lists the corresponding journal article for the model, with link to PubMed page (if exists), as well as the number of citations (from google scholar)
- Ion Type: ion permeability of the channel model; must be one of the following: potassium (K), sodium (Na), calcium (Ca), nonspecific, depending on the Family Selection.
- Subtype: subtype of the ion channel model, as listed in the journal article and/or mod file itself.
- Animal Model: The emulated animal model of the simulation, as either stated explicitly, or inferred from the journal article.
- Brain Area: the emulated brain area and layer of the simulation, as stated explicitly, or inferred from the journal article.
- Neuron Area: The neuron area that the ion channel is found in, divided into dendrites, soma, axon, axon hillock, or specific areas of dendrites and axon.
- Neuron Type: The emulated neuron type of the simulation. May be several types, or general if no neuron type was specified.
- Runtime Q: Elapsed CPU time for running 10 repetitions of a single voltage-clamp protocol (action-potential), divided into four quartiles of the distribution of runtimes of all models in each class.
- Age: animal model age, if specified
- Temp: Details about the model's temperature dependence, and also the temperature at which simulations and/or experiments were performed as described in the journal article.
- Runtime: Elapsed CPU time for running 10 repetitions of a single voltage-clamp protocol (action-potential)
- Comments: Comments from the mod files themselves and any other information about the channel and model from the journal article, such as previous models or experimental data that were used to build the model.
Submit a correction: At the bottom of Channel details, there is a button which can be used to submit a correction (Figure 5). This can be used to report and missing or incorrect information about this channel model. Furthermore, if any information is incorrect that is not related to a specific channel, it can also be listed here. Clicking on the button brings up a new page which allows the user to submit a ticket (Figure 6).
Simulation traces: Just under the channel ID and name, is a clickable tab that shows the simulation traces for the chosen model (Figure 7). Traces are shown for the five simulation protocols: activation, inactivation, deactivation, action potential, and ramp. The first three protocols have multiple voltage step values, as indicated by the colorbars on the right. Note that the entire protocol is not shown - only the relevant part of each protocol is plotted, as described in the corresponding publication.
3: Type Selection
3.1: Hierarchy view
Models are represented as nodes, with edges that represent ancestor-descendant connections (see publication). Groups of models are displayed in dynamically placed arbitrary locations. Hierarchy view has the additional option of adding directional arrowheads to the edges, pointing from ancestor to descendant (Figure 14).
3.2: XY view
Models are represented as nodes and plotted along two metadata dimensions. Note the two extra fields in the available in the toolbar for selection of data.
3.3: Similarity view
Models are represented as nodes, and plotted in PCA coordinate space, according to the first two principal components. Similarity view has the option of adding ancestor-descendant edges, as well as directional arrowheads (Figure 15).
3.4: Cluster view
Models are represented along a circular wheel, in order as defined by their cluster. The outer part of the wheel represents individual models, while the inner wheel represents clusters. Collapsible clustering: In cluster view, clusters have a tooltip (Figure 12), and can also be selected, which collapses the cluster wheel and shows only models in the selected cluster (Figure 13). To return to the full view, click the circle in the middle.
4: Family Selection
Select among the five ion type families: voltage-dependent potassium (K), voltage-dependent sodium (Na), voltage-dependent calcium (Ca), calcium-dependent potassium (KCa), and voltage-dependent hyperpolarization-activated cation (Ih).
5: Color Selection
This allows coloring nodes by any given metadata field, which can be done in all of the views (Figure 16, Figure 17, Figure 18). Legend appears in the bottom right. Note that some colors flicker when there is more than one entry for a given model. Furthermore, for items with greater than 12 possibilities, only the 11 most prevalent entries are shown in color, with all other entries grouped together in the group Other, colored in grey. In cluster view, the clusters themselves are also colored by the most prominent item for the selected metadata category (Figure 18). Color selection options are the same as metadata items listed in Channel Details, but a few extra items are described below:
- Date: Colorbar of publication year, from 1964 to 2014.
- Distance: Color by distance to a selected node. Red indicates small distance and blue indicates large distance.
- Runtime Q: Runtime quantile
- Cluster Representative: reference model for each cluster, defined as the model closest to the center of each cluster in score space.
6: Filter Selection
Filter the models that are currently being shown. We show some examples here. Filters can be added as a positive (Figure 19) or negative (Figure 20) condition. Positive conditions mean that only models with that metadata parameter are shown. A negative filter means that all models without that filter are shown. Any number of positive and negative filters can be added together (Figure 21). Filtering in cluster view does not reduce the number of models being shown, but fades out the models that do not have that metadata attribute (Figure 22). Note also in cluster view that clusters remain visible if they have at least one model with that particular filter attribute.
7: Compare Models
Select models by
shift+drag as shown in Figure 23 and Figure 24. Click the link at the bottom left:
compare X models. This brings you to the compare page (Figure 25). To clear your current selection of models,
shift+click on any whitespace area without a model.
8.1: Example 1: Search for a specific ModelDB entry or author
Searching for ion channel models from a particular modelDB entry can be done in two ways. First of all, models are ordered in the Channels List at the bottom of the right toolbar when no model is selected (Figure 1 B). Alternatively, one could take advantage of the text search filter to enter a modelDB entry, as illustrated in Figure 26 for ModelDB ID 101629. Note that the number of models in other ion type families is displayed in the Filters section as well.
Similarly, when searching for a particular author, the text search filter comes in handy. This is illustrated in Figure 27 for the last name Migliore.
8.2: Example 2: Compare all models of a particular subtype
In this example, let’s say that you wanted to find a m-type potassium model. You would start by filtering the potassium models to only show models with subtype m-type (Figure 28). You would discover that there are 105 m-type models in the database. You then might choose to color the models by cluster, to see what is the diversity of kinetics (Figure 29). You see that there are eight different clusters.
You might be curious how similar different models in the dark blue cluster are to one another, so you instead choose to color by distance. You see that all m-type models in the blue cluster are very similar, nearly identical, and that all m-type models are quite similar (Figure 30). You filter out all clusters except for the dark blue one, to isolate these and then compare them together (Figure 29, Figure 31). Finally, you see the small differences in kinetics between the models, and the differences in metadata between the different models.
You are then interested to compare different m-type models together, so you go back to the browser and select one model from each of the eight different clusters to compare them (Figure 32). You see that there are several unique m-type behaviors, but also that some models do not resemble m-type as much.
Lastly, you might wonder if you are missing some models that might be m-type but not labelled as m-type. In order to find these, you turn to cluster view to find the clusters that are mostly m-type (Figure 33). You select one of these, and can then see several models with subtypes such as HH and LVA that also appear to have m-type behavior. You can then compare these models as well (Figure 34).
8.3: Example 3: Search for a specific brain area and neuron type
In this example, let’s say that you were making a model of a hippocampal interneuron, and you were interested in what ion channel models are available. You would start by filtering the available models down based on brain area and neuron type. You can then color the models by subtype to see that there are a few different subtypes available (Figure 35). You are curious how many behaviors there are for each one, so you go to the XY view and make the axes subtype and cluster (Figure 36). You see that there are seven behaviors for A-type, six behaviors for dr, two for m-type and so on. Let’s say you are interested in looking at the delayed rectifier models, so you select all of these to compare (Figure 37). Just in case some not specified channels are also delayed-rectifier-like, you select these as well, and go to the compare screen. You can now see all of the models available for your particular study and compare them side-by-side according to both metadata and kinetics.