1. Single Phylogeny Example
In this section we demonstrate a simple example of extracting and formatting data from a single phylogeny. We use a simulated phylogeny, which faciliates explaining how the data is structured.
First we simulate a phylogeny using the package ape.
set.seed(
1,
kind = "Mersenne-Twister",
normal.kind = "Inversion",
sample.kind = "Rejection"
)
phylo <- ape::rcoal(10)Important: DAISIEprep requires the tip labels
(taxon names) in the phylogeny to be formatted as genus name and species
name separated by an underscore (e.g. “Canis_lupus”). They can
also optionally have tags appended after the species name (separated by
underscore, e.g. “Canis_lupus_123”; “Canis_lupus_familiaris_123”). This
is common if there are multiple tips in the phylogeny for a single
species, when multiple populations or multiple subspecies have been
sampled. Samples with the same Genus_species name on the tip will be
considered to be of the same species, even if they have subsequent
sampling or subspecies tags.
Here we add tip labels to the simulated phylogeny. In this case, all taxa sampled are different plant species from the same genus.
phylo$tip.label <- c("Plant_a", "Plant_b", "Plant_c", "Plant_d", "Plant_e",
"Plant_f", "Plant_g", "Plant_h", "Plant_i", "Plant_j")Then we convert the phylogeny to a phylo4 class defined
in the package phylobase. This allows users to easily work
with data for each tip in the phylogeny, for example whether they are
endemic to the island or not.
Now we have a phylogeny in the phylo4 format to which we
can easily append data. In this example, we randomly simulate island
endemicity status for each tip, assuming each species has an equal
probability of being not on the island ("not_present"),
endemic to the island ("endemic") or non-endemic to the
island ("nonendemic"). (In a real example this
should be based on the actual endemicity status of each
species!).
endemicity_status <- sample(
x = c("not_present", "endemic", "nonendemic"),
size = length(phylobase::tipLabels(phylo)),
replace = TRUE,
prob = c(0.6, 0.2, 0.2)
)Next, we can add the endemicity data to our phylogenetic tree using
the phylo4d class, again from the phylobase
package. This call is designed for phylogenetic and trait data to be
stored together. The endemicity status needs to be converted into a data
frame in order for the column to be labelled correctly.
We can now visualise our phylogeny with the island endemicity
statuses plotted at the tips. This uses the ggtree and
ggplot2 packages.
Now that we can see the tips that are present on the island, we can
extract them to form our island community data set that can be used in
the DAISIE R package to fit likelihood models of island
colonisation and diversification.
Before we extract species, we will first create an object to store
all of the island colonists’ information. This uses the
island_tbl class introduced in this package
(DAISIEprep). The island_tbl is an S4 class.
This island_tbl object can then easily be converted to a
DAISIE data list using the function create_daisie_data
(more information on this below).
island_tbl <- island_tbl()
island_tbl
#> Class: Island_tbl
#> [1] clade_name status missing_species col_time
#> [5] col_max_age branching_times min_age species
#> [9] clade_type
#> <0 rows> (or 0-length row.names)We can see that this is a object containing an empty data frame. In order to fill this data frame with information on the island colonisation and diversification events we can run:
island_tbl <- extract_island_species(
phylod = phylod,
extraction_method = "min"
)
island_tbl
#> Class: Island_tbl
#> clade_name status missing_species col_time col_max_age branching_times
#> 1 Plant_g nonendemic 0 0.38003405 FALSE NA
#> 2 Plant_i endemic 0 0.04960523 FALSE NA
#> min_age species clade_type
#> 1 NA Plant_g 1
#> 2 NA Plant_i 1The function extract_island_species() is the main
function in DAISIEprep to extract data from the phylogeny.
In the example above, we used the “min” extraction algorithm. The “min”
algorithm extracts island community data with the assumptions of the
DAISIE model (i.e. no back-colonisation from the island to the
mainland), but we recommend using the “asr” algorithm when
back-colonisation events are present in the data (for example, one
species within a large endemic island radiation colonised another island
or mainland). Each row in the island_tbl corresponds to a
separate colonisation of the island. In this case, two colonist
lineages were identified using the ‘min’ extraction algorithm, one
endemic and another non-endemic.
However, if we do not want to use the “min” algorithm, and instead want to extract the most likely colonisations inferred in an ancestral state reconstruction, we need to know the probability of the ancestors of the island species being on the island to determine the time of colonisation. To do this we can fit one of many ancestral state reconstruction methods. Here we use maximum parsimony as it is a simple method that should prove reliable for reconstructing the ancestral species areas (i.e. on the island or not on the island) for most cases. First, we translate our extant species endemicity status to a numeric representation of whether that species is on the island. We add one, as the ancestral state reconstruction method cannot handle zero as a state.
Now we can plot the phylogeny, which this time includes the node labels for the presence/absence on the island in ancestral nodes.
Sidenote: if you are wondering what the probabilities are at each
node and whether this should influence your decision to pick a
preference for island or mainland when the likelihoods for each state
are equal, we can plot the probabilities at the nodes to visualise the
ancestral state reconstruction using
plot_phylod(phylod = phylod, node_pies = TRUE).
Now we can extract island colonisation and diversification times from the phylogeny using the reconstructed ancestral states of island presence/absence.
island_tbl
#> Class: Island_tbl
#> clade_name status missing_species col_time col_max_age branching_times
#> 1 Plant_g endemic 0 0.7648553 FALSE 0.380034....
#> min_age species clade_type
#> 1 NA Plant_g,.... 1As you can see, in this case using the asr
algorithm we find a single colonisation of the island, as can
be seen by the fact that the island_tbl only has one
row.
Now that we have the island_tbl we can convert this to
the DAISIE data list to be used by the DAISIE inference model.
To convert to the DAISIE data list ( i.e. the input data of the
DAISIE inference model) we use create_daisie_data(),
providing the island_tbl as input. We also need to
specify:
- The age of the island or archipelago. Here we use an island age of
one million years (
island_age = 1). - Whether the colonisation times extracted from the phylogenetic data
should be considered precise (
precise_col_time = TRUE). We will not discuss the details of this here, but briefly by setting this toTRUEthe data will tell the DAISIE model that the colonisation times are known without error. Settingprecise_col_time = FALSEwill change tell the DAISIE model that the colonisation time is uncertain and should interpret this as the upper limit to the time of colonisation and integrate over the uncertainty between this point and either the present time or to the first branching point (either speciation or divergence into subspecies). - The number of species in the mainland source pool. Here we set it to
100 (
num_mainland_species = 100). This will be used to calculate the number of species that could have potentially colonised the island but have not. When we refer to the mainland pool, this does not necessarily have to be a continent, it could be a different island if the source of species immigrating to an island is largely from another nearby island (a possible example of this could be Madagascar being the source of species colonising Comoros). This information is used by the DAISIE model to calculate the colonisation rate of the island.
data_list <- create_daisie_data(
data = island_tbl,
island_age = 1,
num_mainland_species = 100,
precise_col_time = TRUE
)Below we show two elements of the DAISIE data list produced. The
first element data_list[[1]] in every DAISIE data list is
the island community metadata, containing the island age and the number
of species in the mainland pool that did not leave descendants on the
island at the present day. This is important information for DAISIE
inference, as it is possible some mainland species colonised the island
but went extinct leaving no trace of their island presence.
Next is the first element containing information on island colonists
(every element data_list[[x]] in the list after the
metadata contains information on individual island colonists). This
contains the name of the colonist, the number of missing species, and
the branching times, which is a vector containing the age of the island,
the colonisation time and the times of any cladogenesis events.
Confusingly, it may be that the branching times vector contains no
branching times: when there are only two numbers in the vector these are
the island age followed by the colonisation time. Then there is the
stac, which stands for status of colonist. This is a number which tells
the DAISIE model how to identify the endemicity and colonisation
uncertainty of the island colonist (these
are explained here if you are interested). Lastly, the type1or2
defines which macroevolutionary regime an island colonist is in. By
macroevolutionary regime we mean the set of rates of colonisation,
speciation and extinction for that colonist. Most applications will
assume all island clades have the same regime and thus all are assigned
type 1. However, if there is a priori expectation that
one clade significantly different from the rest, e.g. the Galápagos
finches amongst the other terrestrial birds of the Galápagos archipelago
this clade can be set to type 2.
data_list[[2]]
#> $colonist_name
#> [1] "Plant_g"
#>
#> $branching_times
#> [1] 1.0000000 0.7648553 0.3800341
#>
#> $stac
#> [1] 2
#>
#> $missing_species
#> [1] 0
#>
#> $type1or2
#> [1] 1This data list is now ready to be used in the DAISIE maximum
likelihood inference model from the R package DAISIE. For
more information on the DAISIE data structures and their application in
the DAISIE models see this vignette
on optimising parameters using DAISIE
