In Albluca/ElPiGraph.R: Elastic principal graph construction

knitr::opts_chunk$set(echo = TRUE)

Once a tree structure has been defined, it is possible to use it to redefine the dimensionality of the data in such way to obtain a structure dependend dimension

Setup

At first we will create a tree ElPiGraph.

library(ElPiGraph.R)
library(igraph)

TreeEPG <- computeElasticPrincipalTree(X = tree_data, NumNodes = 50,
                                       drawAccuracyComplexity = FALSE, drawEnergy = FALSE)

We then need to select a node that will act as a root for the analysis. To do so, we need to visualize the labels of the nodes. To simplify the operation, we will only visualize the id of nodes with degree 1 (i.e., the leaves).

Net <- ConstructGraph(PrintGraph = TreeEPG[[1]])
Nodelab <- 1:vcount(Net)
Nodelab[degree(Net) > 1] <-NA

PlotPG(X = tree_data, TargetPG = TreeEPG[[1]], NodeLabels = Nodelab, LabMult = 8)

For the following example we will use 2 as a the root

BrcStruct <- BranchingDimension(X = tree_data, PrintGraph = TreeEPG[[1]], Start = "2")

The BrcStruct$DimMatrix will now contain the points re-dimensionalised. Each column of the matrix contain a branch dimension between 0 and1

BrcStruct$DimMatrix[is.na(BrcStruct$DimMatrix)] <- 0
pheatmap::pheatmap(BrcStruct$DimMatrix)

Albluca/ElPiGraph.R documentation built on May 28, 2019, 11:02 a.m.