In dtkaplan/DCF: Distributing materials for the DCF course

require(DCFd)
require(ggplot2)
require(dplyr)
require(igraph)

This is an attempt to provide a simple interface for drawing networks using ggplot(). I'm trying to get igraph to do the layout, then let edge properties be set by ggplot.

An example Network

These are connections among NCI60 cell types, using the correlation between probe expression. They came from the cellCor data frame in 2013-08-03-DTK-103.Rmd.

CellEdges <- read.csv("nciNetwork.csv")
CellEdges <- select(CellEdges, corr=correlation, one=cellLine, two=otherCellLine)

What to return: functions edges and pos

geom_graph <- function( Edges, from, to, Vertices, x=NULL,y=NULL ) {

  # Each of these will be a function that draws the 
  # edges and vertices respectively.
  # Use with a plain ggplot() as the base.

  geom_graph <- list( edge_geom, vertex_geom )
  }

}

Find positions for the vertices

The user can hand off a dataframe Pos that specifies the name, x, and y position of each node. If there is no such array, the Edges array will be used to extract the vertex names and assign them a position.

Pos must have only two columns.

For development purposes:

SmallEdges <- head(select(CellEdges,one,two),10)
SmallEdges <- mutate(SmallEdges,one=as.character(one), two=as.character(two))
SmallVertices <- with(SmallEdges,data.frame( vertex.names=unique(c(one,two))))
SmallVertices$width = 1:nrow(SmallVertices)
SmallVertices$x <- rnorm(nrow(SmallVertices))
SmallVertices$y <- rnorm(nrow(SmallVertices))
SmallNet <- graph.data.frame( SmallEdges )

getGraphPos <- function( Edges, Pos=NULL ) {
  JustEdges <- Edges[,c(1,2)]
  origNames <- names(JustEdges)[1:2]
  names(JustEdges) <- c("FROM","TO")
  if ( is.null( Pos ) ) {
    # Construct the locations using igraph
    # make the graph
    Net <- graph.data.frame( JustEdges )
    where <- layout.fruchterman.reingold( Net )
    Pos <- data.frame( x = where[,1], y = where[,2] )
    Pos$name <- vertex.attributes( Net )$name
  }
  # add the position information to the 
  FROM <- inner_join( JustEdges["FROM"], select(Pos,FROM=name,x,y))
  TO <- inner_join( JustEdges["TO"], select(Pos,TO=name,xend=x,yend=y))
  FROM$Group <- 1:nrow(FROM)
  return( cbind(FROM,TO) )
}

Trying to make a dplyr style interface. I want to have the dataframe, then the from and to variables identified.

Return an aes() or a geom_segment() that can be used with segments.

testing <- function(.edges, from, to, 
                    .vertices=NULL, name, x, y, 
                    layout=layout.circle ){
  vars <- structure(as.list(match.call()[-1]), class = "uneval")
  if (missing(from) || missing(to))
    stop("Must provide 'from' and 'to' variable names.")
  From <- eval( vars$from, envir=.edges )
  To   <- eval(   vars$to, envir=.edges )
  Edges <- data.frame( from=From, to=To)
  if ( !is.null(.vertices) ) {
    if (missing(x) || missing(y))
      stop( "Must provide 'name', 'x' and 'y' variable names for vertices.")
    # User has specified the vertices together with x and/or y positions
    X <- eval( vars$x, envir=.vertices )
    Y <- eval( vars$y, envir=.vertices )
    Names <- eval( vars$name, envir=.vertices )
    Pos <- data.frame( from=Names, .x.start = Y, .y.start = Y )
  } else {
    # get the positions automatically
    Net <- graph.data.frame( Edges )
    where <- layout.fruchterman.reingold( Net )
    Pos <- data.frame( x = where[,1], y = where[,2] )
    Pos$name <- vertex.attributes( Net )$name
  }

  positionedEdges <- getGraphPos(Edges,Pos)
  # these should become GEOMS, one for the edges, one for the nodes.
  return(list(edges=Edges,nodes=Pos))
}

vertexPosition <- function(edges) {
  # where should the vertices be located.

  }

dtkaplan/DCF documentation built on May 15, 2019, 4:57 p.m.