Network manipulation

The package contains many utilities to extract information about phylogenetic networks, and to modify networks. Functions that are not exported are more likely to experience breaking changes in future versions, but can be used by prefixing their name with PhyloNetworks. .

Below is a list of the most useful functions. They typically assume a bicombining network, that is, a network in which each hybrid node has exactly 2 parents (never more).

Getting information on a network

overall network information

tiplabels for taxon labels
getroot gives the root node
pairwisetaxondistancematrix for average distances
vcv for the variance-covariance matrix between taxa under a Brownian Motion model along the network, and sharedpathmatrix for the variance-covariance between all nodes (not just leaves)
istimeconsistent: true or false, to know if for all nodes, the various paths from the root to that node have the same length (as expected if length was proportional to time)
getnodeages assuming the network is time-consistent and ultrametric
displayedtrees or majortree to get the displayed trees or major tree, respectively
hardwiredclusters to get all clusters of taxa on a network
biconnectedcomponents and PhyloNetworks.blobinfo give information about the blobs found within a network; also PhyloNetworks.biconnectedcomponent_entrynodes, PhyloNetworks.biconnectedcomponent_exitnodes, and blobdecomposition
treeedgecomponents are the components after removing hybrid edges
checkroot! to check that the graph is a valid semidirected network with a root node in a admissible rooting position

The following functions all compute distances from the root to each node. Their differ in how they handle time inconsistency: when the distance from the root to a node varies across multiple paths from the root to that node.

node information

To learn about nodes and how some might be related or connected, one can use:

PhyloNetworks.isdescendant and PhyloNetworks.isconnected can be used to learn about the relationship between two nodes
hassinglechild
getparent or getparents if the node is a hybrid
getparentminor
getchild or getchildren if the node has more than one child
isrootof

To learn about the edges connected to a given node, one can use:

edge information

To learn about the nodes related connected to a given edge, one can use the following functions:

PhyloNetworks.descendants for the clade ("hardwired cluster") below an edge
getparent
getchild
getpartneredge gives the hybrid partner of an edge, if it is the parent edge of a hybrid node.
isparentof, ischildof can inform whether an edge and node are connected
hardwiredcluster to get the cluster of taxa below a particular edge

Modifying a network

To modify some of the core components of a network:

rootonedge!, rootatnode!: very useful to root with an outgroup, and rotate! to improve plots
nni! to perform a semidirected nearest neighbor interchange
PhyloNetworks.fliphybrid! to flip the direction of a hybrid edge
PhyloNetworks.unzip_canonical! to "unzip" (or zip down) all reticulations, or PhyloNetworks.rezip_canonical! to undo.
setlength! and setlengths! to change the length of one or more edges, and setgamma! to change a hybrid edge inheritance γ

To remove components from a network:

deleteleaf!
deleteaboveLSA!: the "least stable ancestor" may be different from the root
deletehybridthreshold! to simplify a network by deleting edges with small γ's
PhyloNetworks.shrinkedge! to contract an edge
removedegree2nodes! to suppress degree-2 nodes
shrink3cycles! and shrink2cycles! to contract "cycles" of 2 or 3 edges, which deletes 1 reticulation
PhyloNetworks.deletehybridedge! to remove one hybrid edge.

To add components to a network:

To modify some internal attributes, that don't affect the network topology or edge parameters:

To calibrate a network (modify its edge lengths):

calibratefrompairwisedistances!. This documentation has little about calibration so far, but see this tutorial

Comparing two networks

hardwiredclusterdistance: extends the Robinson-Foulds distance. It's a dissimilarity measure on networks: a dissimilarity of 0 does not guarantee that the 2 networks have the same topology in general. But it does if the networks are in some classes (e.g. trees, level-1, tree-child, and others).