If I have a phylogenetic tree for a genus, how can I tell if a certain protein would be a good phylogenetic marker to use if one wanted to use a molecular clock to date the age of some speciation event in that a genus? How can I look for it on a tree?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Speed of Sequence Divergence
You need to use a sequence which diverged only little among most closely related lineages in your tree and diverge quite a lot among most distantly related lineages. In other words, you need a sequence that diverge at a rate that is convenient for your needs.
If you investigate a young tree (a tree which MRCA of all lineages lived recently) you need a sequence that diverge quickly through time. If you investigate an "old tree" (a tree which MRCA of all lineages lived a long time ago) you need a sequence that diverge slowly through time.
Selection
Selection is affecting how fast two sequences diverge. But because selection is specific to the environment experienced by the individuals in each lineages, it can be more difficult to use sequences under selection as a molecular clock (exception of highly conserved selection which are under purifying selection used for very large trees). So we use neutral sequences.
Rate at which mutations fix
Let $u$ be the mutation rate for the sequence of interest and $N$ be the population size. The number of new mutations per year is equal to $2Nu$ (assuming a diploid population). Now, knowing that the sequence is not under selection, the probability of reaching a single mutation to fix (=to reach a frequency of 1 in the population) is $\frac{1}{2N}$ (see this post for a very introduction to why the probability of fixation is inversely proportional to $N$). As a consequence, the rate at which mutations fix is $2 N u \frac{1}{2N}=u$. In other words, the rate of mutation fixation is the mutation rate (for neutral sequences).
Rate at which mutations fix and sequences divergence
The term, sequence divergence is not clearly define. Typically, one would mean "average number of pairwise differences". Assuming that each new mutation that fixes occurred at a new locus (infinite allele model), then the number of pairwise differences between two lineages is the addition of the number of mutations that fixed in each lineage, that is $2u$. Always under the infinite allele model, the probability of the average number of pairwise differences to equal k is given by a Poisson distribution.
Wrapping up
how can I tell if a certain protein would be a good phylogenetic marker
The sequence
- has to be neutral (it makes things easier)
- has to have a mutation rate of the right order of magnitude (it improves accuracy of the estimate and therefore improves statistical power)
