Package 'langevitour'

k-nearest neighbor denoising of a set of points

Description

Reduce noise in a high-dimensional dataset by averaging each point with its nearby neighbors.

Usage

knnDenoise(X, block = rep(1, nrow(X)), k = 30, steps = 2)

Arguments

X	A matrix of numeric data, or something that can be cast to a matrix. Each row represents a point.
block	Optional. A block for each row in X. A factor, or something that can be cast to a factor. Denoising will be performed independently within each block.
k	Number of nearest neighbors to find around each point (including itself).
steps	Number of steps to take along the directed k-nearest neighbor graph. steps=1 uses the k-nearest neighbors, steps=2 uses the k-nearest neighbors and their k-nearest neighbors, etc.

Details

knnDenoise first finds the k-nearest neighbors to each point (including the point itself). Then, for each point, the average is found of the points reachable in steps steps along the directed k-nearest neighbor graph.

Examples

library(palmerpenguins)

completePenguins <- na.omit(penguins[,c(1,3,4,5,6)])

# Dimensions need to be on comparable scales to apply knnDenoise
scaled <- scale(completePenguins[,-1])

denoised <- knnDenoise(scaled)

langevitour(denoised, completePenguins$species, pointSize=2)

---

Langevin Tour

Description

Make a Langevin Tour HTML widget, which can be used to explore high-dimensional numerical datasets.

Usage

langevitour(
  X,
  group = NULL,
  name = NULL,
  center = NULL,
  scale = NULL,
  extraAxes = NULL,
  lineFrom = NULL,
  lineTo = NULL,
  lineColors = NULL,
  axisColors = NULL,
  levelColors = NULL,
  colorVariation = 0.1,
  pointSize = 1,
  subsample = NULL,
  state = NULL,
  width = NULL,
  height = NULL,
  elementId = NULL,
  link = NULL,
  link_filter = TRUE
)

Arguments

X	The data to plot. A matrix of numeric data, or something that can be cast to a matrix. Rows will be shown as points in the widget. Columns are the variables of your data.
group	A group for each row in X, will be used to color points. A factor, or something that can be cast to a factor.
name	A name for each row in X.
center	Center for each variable. If omitted, the column means will be used.
scale	Scale for each variable. Scale +/- center will be the range of guaranteed visible data. If omitted, a reasonable default will be chosen, equal for all variables. (The default is the largest singular value of the centered X times 2.5.)
extraAxes	A matrix with each column defining a projection of interest. The columns of X %*% extraAxes will be presented as extra "variables".
lineFrom	A vector of row numbers. Draw lines starting at these rows.
lineTo	A vector of row numbers. Draw lines ending at these rows.
lineColors	Character vector. A CSS color for each line.
axisColors	Character vector. CSS colors for each variable and then each extra axis.
levelColors	Character vector. CSS colors for each level of group.
colorVariation	Number between 0 and 1. Individual points are given slightly different brightnesses. How strong should this effect be?
pointSize	Point radius in pixels. A single number, or a number for each row in X.
subsample	For speed, randomly subsample down to this many rows.
state	A JSON string, or an object that htmlwidgets will convert to the correct JSON. Initial widget state settings. The state of a widget can be obtained from its "further controls and information" pane. I am not going to guarantee that states will be compatible between versions of langevitour. Hint: Since JSON uses double quotes, surround the string in single quotes.
width	Width of widget in CSS units, for example "700px" or "100%".
height	Height of widget in CSS units, for example "600px" or "75vh".
elementId	An element ID for the widget, see htmlwidgets::createWidget.
link	A SharedData object from the crosstalk package to share selections and filters with other htmlwidgets. The data in this object is not used, just the keys and group name. The rows of link$origData() should correspond to the rows of X.
link_filter	TRUE or FALSE. If using crosstalk, should hiding groups in langevitour also cause them to be filtered in linked widgets?

Details

The only required argument is X, the high-dimensional collection of points. The group argument is also commonly used so that groups of points can be distinguished by color. Further arguments adjust the appearance or provide advanced features.

langevitour will by default not scale variables individually. If you want variables to be individually scaled, use something like scale=apply(X,2,sd)*4. Using the scale argument rather than modifying X directly ensures the plot axes within the widgets retain the original units.

In Javascript, the langevitour object can be obtained using document.getElementById(elementId).langevitour. For example you could have a button that sets the state of a widget using document.getElementById(elementId).langevitour.setState(desiredState).

Value

An htmlwidget object.

Examples

library(palmerpenguins)

completePenguins <- na.omit(penguins[,c(1,3,4,5,6)])
scale <- apply(completePenguins[,-1], 2, sd)*4

langevitour(
    completePenguins[,-1], 
    completePenguins$species, 
    scale=scale, pointSize=2)


# An example setting the widget's initial state
 
langevitour(
    completePenguins[,-1], 
    completePenguins$species, 
    scale=scale, pointSize=2,
    state='{"guideType":"pca","labelInactive":["bill_length_mm"]}')

---

Shiny bindings for langevitour

Description

Output and render functions for using langevitour within Shiny applications and interactive Rmd documents.

Usage

langevitourOutput(outputId, width = "100%", height = "600px")

renderLangevitour(expr, env = parent.frame(), quoted = FALSE)

Arguments

outputId	output variable to read from
width, height	Must be a valid CSS unit (like '100%', '400px', 'auto') or a number, which will be coerced to a string and have 'px' appended.
expr	An expression that generates a langevitour, usually a block of code ending with a call to langevitour()
env	The environment in which to evaluate expr.
quoted	Is expr a quoted expression (with quote())? This is useful if you want to save an expression in a variable.

Examples

library(shiny)
library(palmerpenguins)

completePenguins <- na.omit(penguins[,c(1,3,4,5,6)])
scale <- apply(completePenguins[,-1], 2, sd)*4

ui <- fluidPage(
    sliderInput('zoom', 'Zoom', 0, min=-1, max=1, step=0.1),
    langevitourOutput('widget')
)

server <- function(input,output) { 
    output$widget <- renderLangevitour({
        langevitour(
            completePenguins[,-1], 
            completePenguins$species, 
            scale=scale * 10^input$zoom, pointSize=2)
    })
}

app <- shinyApp(ui, server)

# Use runApp(app) or runGadget(app) to run app.

---

Principal components of scRNA-Seq of mouse brain cells

Description

Single-cell RNA-Seq gene expression of 2,816 mouse brain cells (Zeisel, 2015). The top 10 principal components were produced using the steps in the Bioconductor OSCA workflow.

Usage

data(zeiselPC)

Format

A data frame with 2,816 rows representing brain cells and 11 columns:

type

Cell type.

PC1

Principal component score.

PC2

Principal component score.

PC3

Principal component score.

PC4

Principal component score.

PC5

Principal component score.

PC6

Principal component score.

PC7

Principal component score.

PC8

Principal component score.

PC9

Principal component score.

PC10

Principal component score.

References

Zeisel, A., Muñoz-Manchado, A. B., Codeluppi, S., Lönnerberg, P., La Manno, G., Juréus, A., Marques, S., Munguba, H., He, L., Betsholtz, C., Rolny, C., Castelo-Branco, G., Hjerling-Leffler, J., & Linnarsson, S. (2015). Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq. Science, 347(6226), 1138–1142. doi:10.1126/science.aaa1934

Examples

data(zeiselPC)
langevitour(zeiselPC[,-1], zeiselPC$type)

---