Species tree vs Gene trees during evolutionary analyses: Does it matter?

Question

[ishengomae]

Hi.

Forgive if the question is not in the realm of this site but I've received quite a lot of help here so I'm tempted to ask anyway.

When one wants to do evolutionary analyses involving certain hypothesis testing, as with the case with maximum likelihood analyses of selection forces acting on gene(s), it is usually advised to specify an evolutionary tree. Now when the analysis involves multiple gene families or cases when one is investigating genome wide cases of positive selection some genes will show evolutionary patterns that are not consistent to standard species evolutionary tree.

My question now is does it matter (will it affect the results?) in terms of evolutionary analysis whether the tree used for hypothesis testing is consistent to the species tree or not? I'm eagerly looking forward for an answer as I'm at a point where I don't have rational basis to choose between trees when a gene tree differs from species tree.

Thanks.

[al-ash]

Did you reach any conclusion? If so, please share - this question is still missing an answer supported by some experience/evidence/literature reference.

[ishengomae]

I was able to follow a main suggestion, that is apply a gene tree for my analysis and it turned out that most of my gene trees were concordant with species phylogeny. And there were no problem anyway in publishing the results e.g. https://www.nature.com/articles/ncomms11519.

[Phil S.]

do you mean by species tree, a phylogeny based on whole genomes and a gene tree is based on a subset of the gene repertoire present in the species?

[ishengomae]

Yes, a species tree can also be deduced from other clues including fossil record. A species tree is a general hypothesis about how species are generally believed to have evolved while gene tree is an hypothesis about the evolution of specific gene(s).

For example one mighty want to test which brain-specific genes have undergone positive selection in the lineage leading to human in the context of primate phylogeny. But if you analyze a single gene you may find human being closely related to a macaque (when a standard close relative is a chimp). My issue is to know if it acceptable to force hypothesis testing to conform to the standard phylogeny or one is at liberty to use gene trees.

[Phil S.]

I am by no means an expert on this field but I'd say that if you want to check whether your hypothesis holds true for an arbitrary but fixed set of genes you can use the gene tree even if it is different from the species tree/ the standard phylogeny.

[ishengomae]

Thanks for your thought into this and if you come across a resource that might shed further light into this please do not hesitate.

Answer 1

[MiguelMorard]

Actually it depends on what your goal is. Do you want to study the gene phylogeny or the species phylogeny?

If you're studying one gene, it hasn't to be consistent to the species tree, as it can have a different "history" than the species in itself.

If you want to study the species phylogeny then you'll have to study different genes and you'll have to select them carefully after what you have different solutions as combining them by concatenation or building a supertree from the independent genes trees for example.

Interesting link: (and the complete series)

http://biologos.org/blog/evolution-basics-incomplete-lineage-sorting-and-ancestral-population-sizes

I'm not an expert, but I hope this can be of help.

[ishengomae]

Thanks. In fact my goal is gene centric as I have a certain biological hypothesis which I want to relate at the genetic level. This requires collecting from different species a number of genes involved directly, or indirectly, in the biological process in question and testing for genetic signals in the evolutionary context. My approach is to analyze each gene at a time, so I guess each gene should have its own gene tree. Again thanks.

Answer 2

[Chrispin Chaguza]

I'm not an expert in this field but my opinion is that it would potentially affect the results albeit not too much. Some tools e.g. PAML estimates amount of selection acting in different branches (or lineages) specified by the tree and if the gene is incorrectly placed in a wrong branch the results wouldn't reflect the 'actual' amount of selection acting on that gene. I think the average estimates (across branches) wouldn't change significantly regardless of the branch that contains the gene.

As suggested by others here, Why not try to generate the gene-specific tree and check if it's different from the species tree based on concatenated sequences? If they are similar topologically then I think you don't have to worry that much about all the issues e.g. incomplete lineage sorting, recombination etc that could bias your results. If it's different, wouldn't it be a good idea to just use the species tree as an estimate for the 'true' gene tree? I'm not entirely sure whether this a right approach.

Several mechanisms can confound phylognetic analyses for example genetic recombination or horizontal gene transfers. You should check whether your gene(s) is frequently involved in such processes.

[ishengomae]

Thanks for the (very) helpful reply.