Orthologs vs PFAM

Question

[timothy.kirkwood]

Hello,

If I have a set of orthologous proteins, it is assumed that these will have a shared function due to the maintenance of evolutionary pressure following speciation (as I understand).

If I have a set of proteins that fall within a single PFAM, is the assumption of shared functionality as strong as it would be for the ortholog group? I am guessing not, as the second paragraph of the reference's introduction explicitly says they do not always share function. But then, no assumption is complete, so I imagine there will also be orthologs that do not share function.

I assume there is a degree of uncertainty regarding how high quality the group of orthologs is (e.g. if it is found by RSD or RBH or tree based methods). But with that in mind, would you say the assumption of shared functionality is stronger for the PFAM group or for the orthologous group?

Thanks!

Reference: Nucleic Acids Research, Volume 47, Issue D1, 08 January 2019, Pages D427–D432, https://doi.org/10.1093/nar/gky995

Edit - for anyone interested there's some interesting discussion here - Pfam Based Functional Annotaion

[Mensur Dlakic]

I think when they started Pfam, the idea was to have entries for individual protein families - thus the name. This in turn means that the intent was to have only orthologs as members of each Pfam entry. However, this is really difficult to do without experiments, and HMMs are so sensitive that they inevitably pull in matches to non-orthologs. I think that's why Pfam eventually ended up being a mix of mostly protein families (orthologs) and superfamilies (a mix of ortho-, para- and homologs).

[timothy.kirkwood]

Ah very interesting, thanks for your comment! Would you consider all members of a Pfam (= a protein family, not a superfamily, unless I misunderstand) to be orthologous then? I thought that they were grouped into Pfams according to domain sharing, which doesn't necessarily imply orthology

[Mensur Dlakic]

Pfam is domain-centric, so I was taking about orthology at the domain level. In many cases, especially in prokaryotes, proteins have a single domain, so the domain orthology is the same as orthology in the general sense.

There is a large number of single-domain and single-copy proteins in prokaryotes, which by definition means that all their members are orthologs. Just about all ribosomal proteins, translation factors, single-chain polymerases, etc belong to orthologous Pfam entries.

Answer 1

[lieven.sterck]

I would say stronger for the orthologous groups than in the PFAM groups. The pfam works only on conserved domains, while orthogroups (usually) take the whole sequence into account.

Moreover, it is not because you share a certain domain that you then also share a common function. Most eukaryotic proteins have a multi domain structure and thus the true biological function is a combination of those domains rather than 1 specific domain (though that also exists in eukaryotes). Some domain are very common (kinase domain, DNA binding, ... ) and others are quite specific.

If solely the absence or presence of a domain would determine function (and/or orthology) the problem of orthology defining would have been solved already (which it is not unfortunately)

[timothy.kirkwood]

Perfect, thanks!

Also, this might be worth opening as a new question, but could I get you opinion on the Ortholog conjecture itself? After asking the question above, I found a paper which is essentially saying the assumption that orthologs share function is flawed, which would be slightly worrying! When I digged into it a little there seems to be people on both sides of the fence.

Moses Stamboulian, Rafael F Guerrero, Matthew W Hahn, Predrag Radivojac, The ortholog conjecture revisited: the value of orthologs and paralogs in function prediction, Bioinformatics, Volume 36, Issue Supplement_1, July 2020, Pages i219–i226, https://doi.org/10.1093/bioinformatics/btaa468

[lieven.sterck]

mmyez :)

indeed, if we consider orthology implies shared function that is an assumption/proximisation we made. In most cases this will hold true though, however there are certainly cases where this will not be the case. That is why in "orthology" there is subsection "functional orthologs" (== those that are similar in sequence and share a common function).

Keep in mind that orthology is largely assigned based on sequence similarity, which does makes sense but is in real biology not the only basis of shared function. On the other hand, you can also have a common function for genes not sharing any or much similarity in sequence (convergent evolution == 2 distinct 'solutions' of evolution coming to the same end point).

Do not be alarmed however, as the basic assumption will hold true for most orthologs.What I mention here are exceptions, certainly not the rule, but that also means you can't say that all orthologs share a common function

[Mensur Dlakic]

but that also means you can't say that all orthologs share a common function

The way I always thought about this is that all orthologs do have the same function - that requirement is part of the definition. If they don't have the same function, they are not orthologs, which is where we get into paralogs, general homologs, and so on.

[lieven.sterck]

well, definition wise, those terms have nothing to do with 'function', they merely determine their origin. Orthologs: originated through speciation event, paralogs through duplication event.

Of course a consequence is that for most of the orthologs they will retain the original function in both species but that does not necessarily has to. In one of the species it for instance might evolve to a new function. Those terms do not say anyhing on their evolution since their origin.

Let me stress though, that will only be the case for a small fraction of them, hence why we generally make the assumption they (orthologs) have an identical function. This in contrast to paralogs who often will have a different (sub)function

[Mensur Dlakic]

I think what you are suggesting here - orthologs that have evolved new functions in different species - is not something that can be proven conclusively. By that I mean either we can't prove that their ancestry came by simple speciation (it could have been a duplication followed by a loss of one member), or we can't prove that their functionality is truly different. How does one test many proteins in many species for many different substrates?.

Therefore, for practical purposes we are left with defining orthologs by function when experimental data is available, or by sequence similarity (and a lack of duplicated member) when it isn't.

[lieven.sterck]

not fully agreeing here. Yes of course it is nearly impossible to conclusively prove anything that is evolution related (especially when talking in Mya evolution).

in any case you should not switch around cause and consequence in the whole orthology field. (the paper timothy.kirkwood referenced above illustrates this nicely, here is another one : Nehrt N.L. et al. (2011) Testing the ortholog conjecture with comparative functional genomic data from mammals. PLoS Comput. Biol., 7, e1002073.) )

in any case, you should not use 'function' in any way when making an/the ortholog definition (not even in practise). for the 1-1 orthologs the function consequences are kinda straightforward, but those are only a fraction of all orthologs. It becomes much harder for the n-n orthologs for instance. (let alone we bring in co-orthologs, in/out-paralogs, ... )

[Dunois]

Are there examples of highly similar sequences from different species that aren't orthologs?

[Joe]

I don't have an example off-hand, but this is essentially convergent evolution, so you can read up on this for examples.

[Dunois]

Were there really people on the other side of that fence?

[timothy.kirkwood]

https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1002073 but I double checked and the last authors are the same, so not exactly an independent position :P