Here are some definitions of some key ideas encountered in this documentation.
A “gene tree”, i.e., the genealogical tree describing how a collection of genomes (usually at the tips of the tree) are related to each other at some chromosomal location. See Nodes, Genomes, or Individuals? for discussion of what a “genome” is.
- tree sequence
A “succinct tree sequence” (or tree sequence, for brevity) is an efficient encoding of a sequence of correlated trees, such as one encounters looking at the gene trees along a genome. A tree sequence efficiently captures the structure shared by adjacent trees, (essentially) storing only what differs between them.
Each branching point in each tree is associated with a particular genome in a particular ancestor, called a “node”. Since each node represents a specific genome it has a unique
time, thought of as its birth time, which determines the height of any branching points it is associated with. See Nodes, Genomes, or Individuals? for discussion of what a “node” is.
In certain situations we are interested in how nodes (representing individual homologous genomes) are grouped together into individuals (e.g. two nodes per diploid individual). For example, when we are working with polyploid samples it is useful to associate metadata with a specific individual rather than duplicate this information on the constituent nodes. See Nodes, Genomes, or Individuals? for more discussion on this point.
The focal nodes of a tree sequence, usually thought of as those from which we have obtained data. The specification of these affects various methods: (1)
TreeSequence.haplotypes()will output the genotypes of the samples, and
Tree.rootsonly return roots ancestral to at least one sample. (This can be checked with
is_sample(); see the node table definitions for information on how the sample status a node is encoded in the
The topology of a tree sequence is defined by a set of edges. Each edge is a tuple
(left, right, parent, child), which records a parent-child relationship among a pair of nodes on the on the half-open interval of chromosome
Tree sequences can define the mutational state of nodes as well as their topological relationships. A site is thought of as some position along the genome at which variation occurs. Each site is associated with a unique position and ancestral state.
A mutation records the change of state at a particular site ‘above’ a particular node (more precisely, along the branch between the node in question and its parent). Each mutation is associated with a specific site (which defines the position along the genome), a node (which defines where it occurs within the tree at this position), and a derived state (which defines the mutational state inherited by all nodes in the subtree rooted at the focal node). In more complex situations in which we have back or recurrent mutations, a mutation must also specify its ‘parent’ mutation.
An event at which a parent and child node were born in different populations.
A grouping of nodes, e.g., by sampling location.
An entry recording the origin and history of the data encoded in a tree sequence.
In the set of interconnected tables that we define here, we refer throughout to the IDs of particular entities. The ID of an entity (e.g., a node) is defined by the position of the corresponding row in the table. These positions are zero indexed. For example, if we refer to node with ID zero, this corresponds to the node defined by the first row in the node table.
- sequence length
This value defines the coordinate space in which the edges and site positions are defined. This is most often assumed to be equal to the largest
rightcoordinate in the edge table, but there are situations in which we might wish to specify the sequence length explicitly.
Nodes, Genomes, or Individuals?#
The natural unit of biological analysis is (usually) the individual. However, many organisms we study are diploid, and so each individual contains two homologous copies of the entire genome, separately inherited from the two parental individuals. Since each monoploid copy of the genome is inherited separately, each diploid individual lies at the end of two distinct lineages, and so will be represented by two places in any given genealogical tree. This makes it difficult to precisely discuss tree sequences for diploids, as we have no simple way to refer to the bundle of chromosomes that make up the “copy of the genome inherited from one particular parent”. For this reason, in this documentation we use the non-descriptive term “node” to refer to this concept – and so, a diploid individual is composed of two nodes – although we use the term “genome” at times, for concreteness.
Several properties naturally associated with individuals are in fact assigned
to nodes in what follows: birth time and population. This is for two reasons:
First, since coalescent simulations naturally lack a notion of polyploidy, earlier
tskit lacked the notion of an individual. Second, ancestral
nodes are not naturally grouped together into individuals – we know they must have
existed, but have no way of inferring this grouping, so in fact many nodes in
an empirically-derived tree sequence will not be associated with individuals,
even though their birth times might be inferred.