DeadBird documentation

Model structure

The main object of in this package is the PhyloBDP model, which bundles

a phylogenetic tree
a specification of the duplication, loss and gain rates across the tree
a prior on the number of lineages at the root

In addition, the PhyloBDP model object requires the bound on the number of lineages at the root that leave observed descendants. This bound is determined by the data, and is returned by the functions that read in data in DeadBird.

using DeadBird, NewickTree, DataFrames

First the data side of things:

data = DataFrame(:A=>[1,2], :B=>[0,1], :C=>[3,3])
tree = readnw("((A:1.0,B:1.0):0.5,C:1.5);")
dag, bound = CountDAG(data, tree)

(dag = CountDAG({10, 10} directed simple Int64 graph), bound = 6)

Now we specify the model

rates = ConstantDLG(λ=0.5, μ=0.4, κ=0.1)
prior = ShiftedGeometric(0.9)
model = PhyloBDP(rates, prior, tree, bound)

PhyloBDP(
condition: root
bound: 7 
ConstantDLG{Float64}
  λ: Float64 0.5
  μ: Float64 0.4
  κ: Float64 0.1
)

The model allows likelihood based-inference

using Distributions
loglikelihood(model, dag)

-8.810828109915848

Data structures

There are two main data structures to represent the count data.

(1) There is the CountDAG, which efficiently reduces the data to minimize the required computations when all families (rows) share the same model parameters.

dag, bound = CountDAG(data, tree)

(dag = CountDAG({10, 10} directed simple Int64 graph), bound = 6)

(2) There is the ProfileMatrix, which can be used when model parameters are different across families (rows).

mat, bound = ProfileMatrix(data, tree)

(matrix = Profile{Float64, Int64}[4 1 … 0 3; 6 3 … 1 3], bound = 6)

Both give identical results

loglikelihood(model, dag) == loglikelihood(model, mat)

true

This page was generated using Literate.jl.