dendropy.model.birthdeath: The Birth-Death and Related Processes

Models, modeling and model-fitting of birth-death processes.

  • Nee, S. 2001. Inferring speciation rates from phylogenies. Evolution 55:661-668.
  • Yule, G. U. 1924. A mathematical theory of evolution based on the conclusions of Dr. J. C. Willis. Phil. Trans. R. Soc. Lond. B 213:21-87.
  • Hoehna, S. (2015). The time-dependent reconstructed evolutionary process with a key-role for mass-extinction events. Journal of theoretical biology, 380, 321-331.
dendropy.model.birthdeath.birth_death_likelihood(**kwargs)

Calculates the log-likelihood of a tree (or a set of internal nodes) under a birth death model.

Requires either a Tree object or an interable of internal node ages to be passed in via keyword arguments tree or internal_node_ages, respectively. The former is more convenient when doing one-off calculations, while the latter is more efficient if the list of internal node ages needs to be used in other places and you already have it calculated and want to avoid re-calculating it here.

Parameters:**kwargs (keyword arguments, mandatory) –

Exactly one of the following must be specified:

tree : a Tree object.
A Tree object. The tree needs to be ultrametric for the internal node ages (time from each internal node to the tips) to make sense. The precision by which the ultrametricity is checked can be specified using the ultrametricity_precision keyword argument (see below). If tree is given, then internal_node_ages cannot be given, and vice versa. If tree is not given, then internal_node_ages must be given.
internal_node_ages : iterable (of numerical values)
Iterable of node ages of the internal nodes of a tree, i.e., the list of sum of the edge lengths between each internal node and the tips of the tree. If internal_node_ages is given, then tree cannot be given, and vice versa. If internal_node_ages is not given, then tree must be given.

The following keyword parameters are mandatory:

birth_rate : float
The birth rate.
death_rate : float
The death rate.

The following keyword parameters are optional:

sampling_probability
The probability for a species to be included in the sample. Defaults to 1.0 (all species sampled).
sampling_strategy
The strategy how samples were obtained. Options are: uniform|diversified|age.
is_mrca_included
Does the process start with the most recent common ancestor?
condition_on : string
Do we condition the process on: “time”, “survival”, or “taxa”?

The following are optional, and are only used if internal node ages need to be calculated (i.e., ‘tree’ is passed in).

ultrametricity_precision : float
When calculating the node ages, an error will be raised if the tree in o ultrametric. This error may be due to floating-point or numerical imprecision. You can set the precision of the ultrametricity validation by setting the ultrametricity_precision parameter. E.g., use ultrametricity_precision=0.01 for a more relaxed precision, down to 2 decimal places. Use ultrametricity_precision=False to disable checking of ultrametricity precision.
ignore_likelihood_calculation_failure: bool (default: False)
In some cases (typically, abnormal trees, e.g., 1-tip), the likelihood estimation will fail. In this case a ValueError will be raised. If ignore_likelihood_calculation_failure is True, then the function call will still succeed, with the likelihood set to -inf.

The following are optional, and are only used if internal node ages are specified (i.e., ‘internal_node_ages’ are passed in):

is_node_ages_presorted : bool
By default, the vector of node ages are sorted. If this argument is specified as True, then this sorting will be skipped, in which case it is the client code’s responsibility to make sure that the node ages are given in REVERSE order (i.e., oldest nodes – nodes closer to the root – given first).

Notes

Lifted directly from the (fantastic!) TESS package for R:

H{“o}hna S. 2013. Fast simulation of reconstructed phylogenies under global time-dependent birth–death processes. Bioinformatics, 29(11) 1367-1374.
Returns:
  • lnl (float)
  • The log-likehood of the tree under the birth-death model.
dendropy.model.birthdeath.birth_death_tree(birth_rate, death_rate, birth_rate_sd=0.0, death_rate_sd=0.0, **kwargs)

Returns a birth-death tree with birth rate specified by birth_rate, and death rate specified by death_rate, with edge lengths in continuous (real) units.

Tree growth is controlled by one or more of the following arguments, of which at least one must be specified:

  • If num_extant_tips is given as a keyword argument, tree is grown until the number of EXTANT tips equals this number.
  • If num_extinct_tips is given as a keyword argument, tree is grown until the number of EXTINCT tips equals this number.
  • If num_total_tips is given as a keyword argument, tree is grown until the number of EXTANT plus EXTINCT tips equals this number.
  • If ‘max_time’ is given as a keyword argument, tree is grown for a maximum of max_time.
  • If gsa_ntax is given then the tree will be simulated up to this number of EXTANT tips (or 0 tips), then a tree will be randomly selected from the intervals which corresond to times at which the tree had exactly num_extant_tips leaves. This allows for simulations according to the “General Sampling Approach” of Hartmann et al. (2010). If this option is specified, then num_extant_tips MUST be specified and num_extinct_tips and num_total_tips CANNOT be specified.

If more than one of the above is given, then tree growth will terminate when any one of the termination conditions are met.

Parameters:
  • birth_rate (float) – The birth rate.
  • death_rate (float) – The death rate.
  • birth_rate_sd (float) – The standard deviation of the normally-distributed mutation added to the birth rate as it is inherited by daughter nodes; if 0, birth rate does not evolve on the tree.
  • death_rate_sd (float) – The standard deviation of the normally-distributed mutation added to the death rate as it is inherited by daughter nodes; if 0, death rate does not evolve on the tree.
Keyword Arguments:
 
  • num_extant_tips (int) – If specified, branching process is terminated when number of EXTANT tips equals this number.
  • num_extinct_tips (int) – If specified, branching process is terminated when number of EXTINCT tips equals this number.
  • num_total_tips (int) – If specified, branching process is terminated when number of EXTINCT plus EXTANT tips equals this number.
  • max_time (float) – If specified, branching process is terminated when time reaches or exceeds this value.
  • gsa_ntax (int) – The General Sampling Approach threshold for number of taxa. See above for details.
  • tree (Tree instance) – If given, then this tree will be used; otherwise a new one will be created.
  • taxon_namespace (TaxonNamespace instance) – If given, then this will be assigned to the new tree, and, in addition, taxa assigned to tips will be sourced from or otherwise created with reference to this.
  • is_assign_extant_taxa (bool [default: True]) – If False, then taxa will not be assigned to extant tips. If True (default), then taxa will be assigned to extant tips. Taxa will be assigned from the specified taxon_namespace or tree.taxon_namespace. If the number of taxa required exceeds the number of taxa existing in the taxon namespace, new Taxon objects will be created as needed and added to the taxon namespace.
  • is_assign_extinct_taxa (bool [default: True]) – If False, then taxa will not be assigned to extant tips. If True (default), then taxa will be assigned to extant tips. Taxa will be assigned from the specified taxon_namespace or tree.taxon_namespace. If the number of taxa required exceeds the number of taxa existing in the taxon namespace, new Taxon objects will be created as needed and added to the taxon namespace. Note that this option only makes sense if extinct tips are retained (specified via ‘is_retain_extinct_tips’ option), and will otherwise be ignored.
  • is_add_extinct_attr (bool [default: True]) – If True (default), add an boolean attribute indicating whether or not a node is an extinct tip or not. False will skip this. Name of attribute is set by ‘extinct_attr_name’ argument, defaulting to ‘is_extinct’. Note that this option only makes sense if extinct tips are retained (specified via ‘is_retain_extinct_tips’ option), and will otherwise be ignored.
  • extinct_attr_name (str [default: 'is_extinct']) – Name of attribute to add to nodes indicating whether or not tip is extinct. Note that this option only makes sense if extinct tips are retained (specified via ‘is_retain_extinct_tips’ option), and will otherwise be ignored.
  • is_retain_extinct_tips (bool [default: False]) – If True, extinct tips will be retained on tree. Defaults to False: extinct lineages removed from tree.
  • repeat_until_success (bool [default: True]) – Under some conditions, it is possible for all lineages on a tree to go extinct. In this case, if this argument is given as True (the default), then a new branching process is initiated. If False (default), then a TreeSimTotalExtinctionException is raised.
  • rng (random.Random() or equivalent instance) – A Random() object or equivalent can be passed using the rng keyword; otherwise GLOBAL_RNG is used.

References

Hartmann, Wong, and Stadler “Sampling Trees from Evolutionary Models” Systematic Biology. 2010. 59(4). 465-476

dendropy.model.birthdeath.discrete_birth_death_tree(birth_rate, death_rate, birth_rate_sd=0.0, death_rate_sd=0.0, **kwargs)

Returns a birth-death tree with birth rate specified by birth_rate, and death rate specified by death_rate, with edge lengths in discrete (integer) units.

birth_rate_sd is the standard deviation of the normally-distributed mutation added to the birth rate as it is inherited by daughter nodes; if 0, birth rate does not evolve on the tree.

death_rate_sd is the standard deviation of the normally-distributed mutation added to the death rate as it is inherited by daughter nodes; if 0, death rate does not evolve on the tree.

Tree growth is controlled by one or more of the following arguments, of which at least one must be specified:

  • If ntax is given as a keyword argument, tree is grown until the number of tips == ntax.
  • If taxon_namespace is given as a keyword argument, tree is grown until the number of tips == len(taxon_namespace), and the taxa are assigned randomly to the tips.
  • If ‘max_time’ is given as a keyword argument, tree is grown for max_time number of generations.

If more than one of the above is given, then tree growth will terminate when any of the termination conditions (i.e., number of tips == ntax, or number of tips == len(taxon_namespace) or number of generations = max_time) are met.

Also accepts a Tree object (with valid branch lengths) as an argument passed using the keyword tree: if given, then this tree will be used; otherwise a new one will be created.

If assign_taxa is False, then taxa will not be assigned to the tips; otherwise (default), taxa will be assigned. If taxon_namespace is given (tree.taxon_namespace, if tree is given), and the final number of tips on the tree after the termination condition is reached is less then the number of taxa in taxon_namespace (as will be the case, for example, when ntax < len(taxon_namespace)), then a random subset of taxa in taxon_namespace will be assigned to the tips of tree. If the number of tips is more than the number of taxa in the taxon_namespace, new Taxon objects will be created and added to the taxon_namespace if the keyword argument create_required_taxa is not given as False.

Under some conditions, it is possible for all lineages on a tree to go extinct. In this case, if the keyword argument repeat_until_success is True, then a new branching process is initiated. If False (default), then a TreeSimTotalExtinctionException is raised.

A Random() object or equivalent can be passed using the rng keyword; otherwise GLOBAL_RNG is used.

dendropy.model.birthdeath.fit_pure_birth_model(**kwargs)

Calculates the maximum-likelihood estimate of the birth rate of a set of internal node ages under a Yule (pure-birth) model.

Requires either a Tree object or an interable of internal node ages to be passed in via keyword arguments tree or internal_node_ages, respectively. The former is more convenient when doing one-off calculations, while the latter is more efficient if the list of internal node ages needs to be used in other places and you already have it calculated and want to avoid re-calculating it here.

Parameters:**kwargs (keyword arguments, mandatory) –

Exactly one of the following must be specified:

tree : a Tree object.
A Tree object. The tree needs to be ultrametric for the internal node ages (time from each internal node to the tips) to make sense. The precision by which the ultrametricity is checked can be specified using the ultrametricity_precision keyword argument (see below). If tree is given, then internal_node_ages cannot be given, and vice versa. If tree is not given, then internal_node_ages must be given.
internal_node_ages : iterable (of numerical values)
Iterable of node ages of the internal nodes of a tree, i.e., the list of sum of the edge lengths between each internal node and the tips of the tree. If internal_node_ages is given, then tree cannot be given, and vice versa. If internal_node_ages is not given, then tree must be given.

The following are optional, and are only used if internal node ages are specified (i.e., ‘internal_node_ages’ are passed in):

is_node_ages_presorted : bool
By default, the vector of node ages are sorted. If this argument is specified as True, then this sorting will be skipped, in which case it is the client code’s responsibility to make sure that the node ages are given in REVERSE order (i.e., oldest nodes – nodes closer to the root – given first).
Returns:
  • m (dictionary)
  • A dictionary with keys being parameter names and values being
  • estimates –
    “birth_rate”
    The birth rate.
    ”log_likelihood”
    The log-likelihood of the model and given birth rate.

Examples

Birth rates can be estimated by passing in trees directly:

for idx, tree in enumerate(trees):
    m = birthdeath.fit_pure_birth_model(tree=tree)
    print("Tree {}: birth rate = {} (logL = {})".format(
        idx+1, m["birth_rate"], m["log_likelihood"]))

Or by pre-calculating and passing in a list of node ages:

for idx, tree in enumerate(trees):
    m = birthdeath.fit_pure_birth_model(
            internal_node_ages=tree.internal_node_ages())
    print("Tree {}: birth rate = {} (logL = {})".format(
        idx+1, m["birth_rate"], m["log_likelihood"]))

Notes

Adapted from the laser package for R:

dendropy.model.birthdeath.fit_pure_birth_model_to_tree(tree, ultrametricity_precision=1e-05)

Calculates the maximum-likelihood estimate of the birth rate a tree under a Yule (pure-birth) model.

Parameters:tree (Tree object) – A tree to be fitted.
Returns:
  • m (dictionary)
  • A dictionary with keys being parameter names and values being
  • estimates –
    • “birth_rate” The birth rate.
    • ”log_likelihood” The log-likelihood of the model and given birth rate.

Examples

import dendropy
from dendropy.model import birthdeath
trees = dendropy.TreeList.get_from_path(
        "pythonidae.nex", "nexus")
for idx, tree in enumerate(trees):
    m = birthdeath.fit_pure_birth_model_to_tree(tree)
    print("Tree {}: birth rate = {} (logL = {})".format(
        idx+1, m["birth_rate"], m["log_likelihood"]))
dendropy.model.birthdeath.uniform_pure_birth_tree(taxon_namespace, birth_rate=1.0, rng=None)

Generates a uniform-rate pure-birth process tree.