Point¶

Point mark displays each data item as a symbol. Points are often used to create a scatter plot. In the genomic context, they could represent, for example, point mutations at genomic loci.

{
  "data": { "url": "sincos.csv" },
  "mark": "point",
  "encoding": {
    "x": { "field": "x", "type": "quantitative" },
    "y": { "field": "sin", "type": "quantitative" },
    "size": { "field": "x", "type": "quantitative" }
  }
}

Channels¶

In addition to standard position channels and color, opacity, and strokeWidth channels, point mark has the following channels:

size

Type: Number

The area of the point in pixels. In practice, the area is less because the shapes do not fill their rectangular container. Example: the diameter of a circle with the size of 100 is 10 (sqrt(100)) pixels.

Default value: 100

shape

Type: String

One of "circle", "square", "cross", "diamond", "triangle-up", "triangle-down", "triangle-right", or "triangle-left".

Default value: "circle"

dx

Type: Number

The horizontal offset between the text and its anchor point, in pixels. Applied after the rotation by angle.

Default value: 0

dy

Type: Number

The vertical offset between the text and its anchor point, in pixels. Applied after the rotation by angle.

Default value: 0

Properties¶

fillGradientStrength

Type: Number

Gradient strength controls the amount of the gradient eye-candy effect. Valid values are between 0 and 1.

Default value: 0

geometricZoomBound

Type: Number

Enables geometric zooming. The value is the base two logarithmic zoom level where the maximum point size is reached.

Default value: 0

sampleFacetPadding

Type: Number

A special property for the GenomeSpy app.

Padding between sample facet's upper/lower edge and the maximum point size. This property controls how tightly points are squeezed when facet's height is smaller than the maximum point size. The unit is a proportion of facet's height. The value must be between 0 and 0.5. This property has no effect when sample faceting is not used.

Default value: 0.1

Examples¶

Plenty of points¶

The example below demonstrates how points can be varied by using shape, fill, size, strokeWidth, and angle channels.

{
  "data": {
    "sequence": { "start": 0, "stop": 160, "as": "z" }
  },

  "transform": [
    { "type": "formula", "expr": "datum.z % 20", "as": "x" },
    { "type": "formula", "expr": "floor(datum.z / 20)", "as": "y" }
  ],

  "mark": {
    "type": "point",
    "stroke": "black"
  },

  "encoding": {
    "x": { "field": "x", "type": "ordinal", "axis": null },
    "y": { "field": "y", "type": "ordinal", "axis": null },
    "shape": { "field": "x", "type": "nominal" },
    "fill": { "field": "x", "type": "nominal" },
    "size": {
      "field": "x",
      "type": "quantitative",
      "scale": { "type": "pow", "exponent": 2, "range": [0, 900] }
    },
    "strokeWidth": {
      "field": "y",
      "type": "quantitative",
      "scale": { "range": [0, 4] }
    },
    "angle": {
      "field": "y",
      "type": "quantitative",
      "scale": { "range": [0, 45] }
    }
  }
}

Zoom behavior¶

Although points are infinitely small on the real number line, they have a specific diameter on the screen. Thus, closely located points tend to overlap each other. Decreasing the point size reduces the probability of overlap, but in a zoomed-in view, the plot may become overly sparse.

To control overplotting, the point mark provides two zooming behaviors that adjust the point size and visibility based on the zoom level.

Geometric zoom¶

Geometric zoom scales the point size down if the current zoom level is lower than the specified level (bound). geometricZoomBound mark property enables geometric zooming. The value is the negative base two logarithm of the relative width of the visible domain. Example: 0: (the default) full-size points are always shown, 1: when a half of the domain is visible, 2: when a quarter is visible, and so on.

The example below displays 200 000 semi-randomly generated points. The points reach their full size when 1 / 2^10.5 of the domain is visible, which equals about 1500X zoom.

{
  "data": {
    "sequence": { "start": 0, "stop": 200000, "as": "x" }
  },
  "transform": [
    { "type": "formula", "expr": "random() * 0.682", "as": "u" },
    {
      "type": "formula",
      "expr": "((datum.u % 1e-8 > 5e-9 ? 1 : -1) * (sqrt(-log(max(1e-9, datum.u))) - 0.618)) * 1.618 + sin(datum.x / 10000)",
      "as": "y"
    }
  ],
  "mark": {
    "type": "point",
    "geometricZoomBound": 10.5
  },
  "encoding": {
    "x": { "field": "x", "type": "quantitative", "scale": { "zoom": true } },
    "y": { "field": "y", "type": "quantitative" },
    "size": { "value": 200 },
    "opacity": { "value": 0.6 }
  }
}

Tip

You can use geometric zoom to improve rendering performance. Smaller points are faster to render than large points.

Semantic zoom¶

The score-based semantic zoom adjusts the point visibility by coupling a score threshold to current zoom level. The semanticScore channel enables the semantic zoom and specifies the score field. The semanticZoomFraction property controls the fraction of data items to show in the fully zoomed-out view, i.e., it specifies the threshold score. The fraction is scaled as the viewport is zoomed. Thus, if the data is distributed roughly uniformly along the zoomed axis, roughly constant number of points are visible at all zoom levels. The score can be arbitrarily distributed, as the threshold is computed using p-quantiles.

The example below has 200 000 semi-randomly generated points with an exponentially distributed score. As the view is zoomed in, new points appear. Their number in the viewport stays approximately constant until the lowest possible score has been reached.

{
  "data": {
    "sequence": { "start": 0, "stop": 200000, "as": "x" }
  },
  "transform": [
    { "type": "formula", "expr": "random() * 0.682", "as": "u" },
    {
      "type": "formula",
      "expr": "((datum.u % 1e-8 > 5e-9 ? 1 : -1) * (sqrt(-log(max(1e-9, datum.u))) - 0.618)) * 1.618",
      "as": "y"
    },
    {
      "type": "formula",
      "expr": "-log(random())",
      "as": "score"
    }
  ],
  "mark": {
    "type": "point",
    "semanticZoomFraction": 0.002
  },
  "encoding": {
    "x": { "field": "x", "type": "quantitative", "scale": { "zoom": true } },
    "y": { "field": "y", "type": "quantitative" },
    "opacity": {
      "field": "score",
      "type": "quantitative",
      "scale": { "range": [0.1, 1] }
    },
    "semanticScore": { "field": "score", "type": "quantitative" },
    "size": { "value": 100 }
  }
}

Tip

The score-based semantic zoom is great for filtering point mutations and indels that are scored using CADD, for example.