Working with Gates

Gates and Populations in CellEngine

Basics

Gates are geometric shapes that define boundaries within which events (cells) must be contained to be considered part of a population. Gates may be any one of RectangleGate, PolygonGate, EllipseGate, RangeGate, QuadrantGate or SplitGate. Each gate has a gid value, explained below.

On a biological basis, populations are groups of cells that are analyzed together. For example, you might have a “CD4+ T cells” population. In CellEngine’s implementation, populations are gates combined with Boolean operators. For example, “CD4+ T cells” might be defined as {$and: [<Singlets>, <T cells>, <CD4+>]}, where the values in <brackets> are the gid values for those gates. Any of $and, $or, $not and $xor can be used, but note that the CellEngine UI cannot display certain complex populations.

Populations have two additional parameters of note besides their gates:

parentId - This is used to construct the hierarchy.

terminalGateGid - This is an optional parameter that, in a typical, step-wise hierarchy, indicates the one gate that distinguishes a population from its parent. For example, given the hierarchy Ungated > Singlets > T cells, the populations would look like this:

{
  name: "Ungated",
  _id: null,
  gates: '{"$and": []}',
  terminalGateGid: null,
  parentId: null
}

{
  name: "Singlets",
  _id: "64daccf0ff756d82e41155ba",
  gates: '{"$and": ["<singlets GID>"]}',
  terminalGateGid: "<singlets GID>",
  parentId: null
}

{
  name: "T cells",
  _id: "64daccf0ff756d82e41155bb",
  gates: '{"$and": ["<singlets GID>", "<T cells GID>"]}',
  terminalGateGid: "<T cells GID>",
  parentId: "64daccf0ff756d82e41155ba"
}

Boolean populations that don’t have a single gate that distinguishes them from their parent have this value set to NULL.

Gate tailoring

CellEngine allows each FCS file to have a unique gate geometry by “tailoring” gates per-file. Tailored gates exist in a group. Within a group, there’s the global gate, which is used for any FCS file that does not have a file-specific gate; and there are zero or more tailored gates, each of which is used for a specific FCS file. All gates in the group have the same gid (group ID) value. The global gate has an fcsFileId value of NA, and the file-specific gates have fcsFileId values set to the _id of their corresponding FCS file.

Compound gates (quadrant and split) are made up of “sectors.” Quadrant gates have four sectors (upper-right, upper-left, lower-left, lower-right) and split gates have two sectors (left and right). In addition to the top-level gid (like simple gates), these gates have model.gids and names lists that specify the group ID and name for each sector, in the order shown above. Populations using compound gates must reference these sector group IDs.

The above is all you need to know to get started with creating gates with the CellEngine R Toolkit, but the CellEngine API documentation for gates and populations has more detailed information and examples.

Creating a simple gate and population

First let’s create a basic polygon gate.

library("cellengine")
authenticate() # prompts for username and password

# Get an existing experiment called "My experiment"
experiment = getExperiment(byName("My experiment"))

res = createPolygonGate(
  experimentId=experiment$`_id`,
  "FSC-A", "FSC-H", # the X and Y channels
  "my gate", # the gate name
  list( # the vertices
    c(37836.07, 971.51),
    c(1588732.12, 154.646),
    c(8139.405, 664.78),
    c(9441.949, 781.32)
  ),
  createPopulation = TRUE,
  parentPopulationId = cellengine::UNGATED
)

Two things to observe:

We used createPopulation = TRUE and parentPopulationId = cellengine::UNGATED. This tells CellEngine to create a corresponding population for the gate at the root level, under “Ungated”. The res object contains res$gate and res$population members, so you can access both the created gate and created population.
The gate coordinates are in “scaled data space.” That means that the scale function that is used to show the data in CellEngine (i.e., linear, log or arcsinh) also needs to be applied to the gate coordinates. You can see what scale is in used by clicking “edit scales” in CellEngine, or by calling getScaleSets() in the R toolkit.

You can use the toolkit’s applyScale() function to convert from raw data space to scaled data space.

To create the next gate and population in the hierarchy, below “my gate”, repeat the above, but set parentPopulationId = byName("my gate") or parentPopulationId = res$population$`_id`.

Creating a set of tailored gates

First we’re going to make the global gate. This is the same as above, except we’ll add tailoredPerFile = TRUE:

res = createPolygonGate(
  experimentId=experiment$`_id`,
  "FSC-A", "FSC-H",
  "my other gate",
  list(
    c(37836.07, 971.51),
    c(1588732.12, 154.646),
    c(8139.405, 664.78),
    c(9441.949, 781.32)
  ),
  createPopulation = TRUE,
  parentPopulationId = byName("singlets"),
  tailoredPerFile = TRUE # <-- added
)

Now we’re going to make a tailored gate:

res2 = createPolygonGate(
  experimentId=experiment$`_id`,
  "FSC-A", "FSC-H",
  "my other gate",
  list(
    c(12345, 900),
    c(160002, 175),
    c(8000, 750),
    c(9410, 800)
  ),
  gid = res$gate$gid, # <-- added
  tailoredPerFile = TRUE,
  fcsFileId = byName("file1.fcs") # <-- added
  # createPopulation is FALSE or omitted
  # parentPopulationId is omitted
)

Notice:

We set gid = res$gate$gid so that this new gate will be in the same gate group as the gate we created in the previous step.
We added fcsFileId = byName("file1.fcs"). This creates a gate specific to that file.
We removed createPopulation and parentPopulationId because we’re not creating a population again; we’re just creating a gate tailored to this file.
The gate vertices are different from the global gate.
The channels and name are the same as the global gate. CellEngine currently doesn’t strictly require that tailored gates have the same name and channels; it’s up to the user to ensure the same values are used.

Now if you open CellEngine and flip through the FCS files in this experiment, you should see the file-specific gates being used.

Converting to and from flowCore gates

CellEngine has support for converting ellipse, polygon and rectangle gates to and from flowCore gates. More gate types will be supported eventually.

Given a flowCore gate:

rectGate <- rectangleGate(
  filterId = "Gate 1",
  "FL1-H" = c(0, 12), "FL2-H" = c(0, 12)
)

We can add it to a CellEngine experiment as follows:

fromFlowCore(rectGate, experimentId = byName("my experiment"), name = "my gate")
# "my gate" should now exist in the CellEngine experiment.

fromFlowCore() will pass ... arguments on to the create__Gate() function, so you can, for example, create a population by passing createPopulation = TRUE, parentPopulationId = byName("parent"), or tailor the gate by passing tailoredPerFile = TRUE, fcsFileId = byName("myfile.fcs").

Similarly, you can convert a CellEngine gate to flowCore:

ceGate <- getGate(experimentId = byName("my experiment"), gateId = byName("singlets"))
fcGate <- toFlowCore(ceGate)

Creating complex populations

Using createPopulation = TRUE works to create common, step-wise hierarchies. What if you want to make a “not” or “or” population?

# Given GIDs from the gate creation procedures above:
singletsGid = "64daccf0ff756d82e4115588" # the Singlets gate
cd4Gid = "64daccf0ff756d82e4115599" # the CD4+ gate
il2gid = "64daccf0ff756d82e41155aa" # the IL2+ gate
il4gid = "64daccf0ff756d82e41155ab" # the IL4+ gate
# Create an "or(IL2+, IL4+)" population below Singlets > CD4+:
createPopulation(
  experimentId = byName("my experiment"),
  name = "or(IL2+, IL4+)",
  gates = sprintf('{"$and": ["%s", "%s", {"$or": ["%s", "%s"]}]}',
                  singletsGid, cd4Gid, il2gid, il4gid),
  terminalGateGid = NULL,
  parentId = byName("CD4+")
)