Simply comparing the two doesn't give you any _evolutionary_ information at all, just similarity.

It's important to note that similarity is only == relatedness in the case of neutrally-evolving sites at strictly orthologous loci. Where gene duplications and or adaptive evolution (or any other non-neutral process) take place, similar sequences may not be closely related; and dissimilar sequences may be.

I suspect what you're really trying to do is establish whether the two proteins are orthologous in any meaningful sense, i.e. whether they share a common ancestry.

To do this you'd need to implicitly or explicitly take into account *all* the other protiens, in all other taxa, that are vaugely similar. You can use similarity-based algorithms, e.g. reciprocal BLAST, to identify these and then something like an HMMs to model their relationship; or align them and construct a phylogeny or network.

For any kind of comparative phylogenetics you need more than two taxa/sequences, and just comparing differences is meaningless.

I'm not sure I understand the reasoning behind Alessio and Sezer's suggestions.

you need to test substitution models (model test, as for instance in Mega5) and choose the top one. Try also with Bayesian and see posterior probabilities at nodes

What is "similar"? An identity of 35% of more over the full length of the proteins is a good indicator of shared ancestry. Below that it is difficult to tell anything with primary sequence analysis alone.

If you have identity levels above 35%, what I recomend to do is to fish for similar sequences, using PSI-BLAST, or making a HMM with the sequences you currently have: then use that model to screen the deduced proteomes of species of interest, paying attention on defining a good score or e-value threshold to separate true positives from false positive keeping the false negative rate low.

Then with these identified proteins make a phylogenetic analysis, i.e., alignment of the protein sequences (Mafft, muscle), selection of evolutionary model (ProtTest), phylogenetic analysis (RAxML, MrBayes), this COULD help narrowing down their evolutionary history.

All this assumes that the proteins share identity over their full lengths, if that is not the case, and you can only recognize identity over some domains, then the best would be to restrict the analyses to these shared domains.

I Hope that helps.

Depending on what you want to do, two sequences are not enough. For example the suggestion to estimate a substitution model requires a certain minimum of data. Two sequences will not be able to give you a good estimate of the substitution model.

For any kind of phylogenetic analysis, two sequences will not give interesting results.

You could identify domains, e.g. with pfam_scan. Common domains could be a good start even if sequence identity is low.

Diego suggested to add more similar sequences. This could help to put your two sequence in the context of other sequences which might help to identify their mutual relationship. But remember: If you have ortholog and paralog genes in one data set you can get misleading results.

Simply comparing the two doesn't give you any _evolutionary_ information at all, just similarity.

It's important to note that similarity is only == relatedness in the case of neutrally-evolving sites at strictly orthologous loci. Where gene duplications and or adaptive evolution (or any other non-neutral process) take place, similar sequences may not be closely related; and dissimilar sequences may be.

I suspect what you're really trying to do is establish whether the two proteins are orthologous in any meaningful sense, i.e. whether they share a common ancestry.

To do this you'd need to implicitly or explicitly take into account *all* the other protiens, in all other taxa, that are vaugely similar. You can use similarity-based algorithms, e.g. reciprocal BLAST, to identify these and then something like an HMMs to model their relationship; or align them and construct a phylogeny or network.

For any kind of comparative phylogenetics you need more than two taxa/sequences, and just comparing differences is meaningless.

I'm not sure I understand the reasoning behind Alessio and Sezer's suggestions.

For your two protein sequences you should test whether they can be assumed to be homologous. If they are homologous indeed, there can be two possibilities: (1) either they have evolved from one common ancestor, or (2) they have arisen via convergent evolution. To check this, you need to first assemble as many homologous sequences as possible. Above several people already suggested using PSI-BLAST and profile HMMs (eg, use HMMER) to do this. Once you have a full set of homologs, you should use phylogenetic methods to study the evolutionary origin of your two sequences. What hypotheses (1 or 2) does your data support? Here the correctness of the evolutionary model you choose for phylogeny analysis is important. In general I would use PhyML and LG model for this, unless your data in some way special.

Hope this helps :-)

There are many proteins, both enzymes and structural proteins, which evolved before archeabacteria, eubacteria and eukaryotes split from the common ancestor. Many more are shared among all eukaryotes (plants, animals, fungi, protists). Not all that are shared among all (or most, perhaps being lost in some lineages) eukaryotes were present in the common ancestor of the eukaryotes, the alternative is that they were horizontally transferred. Horizontal transfer between plants and animals is rather rare but not impossible. Convergent evolution to a similar enzyme function from non homologous origins (homolog means shared a common origin, it is not a symonym of similarity) is common, but most often does not result in similar sequences.

If your enzyme shared an ancestor before plants split from animals, you may also be able to find it in fungi, protists, etc... If it was horizontally transferred, you may be able to find evidence for that. There is no specific "cut off" for what percentage of identity between the sequences "proves" homology vs convergent evolution. But if the protein is highly conserved in all animals (vertebrates, invertebrates, etc) it would not be unexpected to be able to also find it in plants and other non-animal eukaryotes.

Similarity of sequences is an good indicator for a common ancestor of your two sequences. Question is how you measure it: score; similarity; e-value. For the latter you need a database and a eg BLAST query. Second step is to validate the Result.

This is done by protein family inference approaches, which one compared to another depend of the input: set of species; set of proteins etc. All the existing PF approaches differ in general in this second point. Not knowing your sequences and your species, you can have a look into a PF approach that cover the universal protein space. I know that http://systers.molgen.mpg.de does this. Find out the Protein Families (sequence clusters) for both proteins and inspect sequences and species. Are both of your sequences in the SYSTERS set and further in the same family then this is a good indicator that evolutionary relationship is given even if SYSTERS only claims for to cluster similar sequences into a family (see my first sentence). Do you find only similar sequences, consider similarit of your best hit. You can balst against the family consensus in SYSTERs. Compare to other family approaches; use ready aproaches; see that I argue with the arguments of the other before my contribution. Good success!

Phylogenetic analysis of nucleotide sequences of th eencodin genes for the two enzymes complemented with a suitable range (depending on evolutionary rate and evolutionary distance) of other homologous sequences.

...by the way, BLAST similarity scores in NOT (I repeat NOT) a measure of common evolutionary history... unlike what is seen in some contemporary papers.

Right, Anders. The second step is to validate these scores (e-values) in most of protein family inferring aproaches. Here, the initiators have crated the algorithm or a complete pipeline to fulfill this task. The interpretation of the result is evolutionary relationship (or none).

I agree with, Thomas that similarity of sequences is a good indicator for a common ancestor of your two sequences. The crux of it too is would two sequences be just compared in a 'one sample T-Test' sort of arrangement in genomics and arrive at an evolutionary commonality conclusion? I guess that is perhaps what is at the back of Bhaskar's way of looking at it. I guess there is a large possibility support for this under the use of such computational systems such as SYSTERS.

There seems to be a contradiction in Maria's answer. The way I understand this, homologous proteins cannot come from convergent evolution. Perhaps what you meant in the first sentence was similarity. It is important to keep a clear distinction between similarity and homology, precisely to understand that convergent evolution can lead to sequence similarity, but homology is excluded by definition (http://www.britannica.com/EBchecked/topic/270557/homology).

there are some points you need to consider:

- are these two protein molecules enzymes or one of them is and the other is proenzyme or drived protein?

- do they catalyze the same reaction?

- do they have the same working environments of pH, Temp, ionic strength...ect?

- if they catalyze the same reaction, do they give the same product?

the presence of homology does not mean that the active moiety is the same in both enzymes.

I would consider points raised by Ashraf Khodeir.

It is possible that the protein/enzyme you are investigating is an important factor imperative for sustaining as a biological life form. If it is so... its function must have been essentially present throughout evolution. You should check if orthology exist between these two protein/enzyme. You can do this by searching databases like COG, OrthoMCL and Database of Essential Genes(DEG).

I'm surprised no one has suggested looking at crystal structures if there are any. If a crystal structure of the exact enzyme doesn't exist, a close homolog (%id > 50%) should suffice. Homologous proteins generally have a similar arrangement of secondary structural elements and this will persist even when the sequence identity is well into the random range.

You mention that both proteins are enzymes. The first question should be whether they carry out similar reactions. In the evolution of enzymes, the details of the chemistry are most highly conserved, for instance a nucleophile or metal-binding site (where that metal is involved in the chemistry). As a rough approximation the first few digits of the EC (Enzyme Commission) code can be used as a guide, but note that the EC number can change depending on which direction of the reaction is referenced (e.g., hydrolysis vs ligation).

So if the proteins have similar folds (see for example the CATH database) and if they carry out similar reactions with similar key active site residues (e.g. a nucleophilic serine or cysteine, or an active-site metal binding histidine) then you can make a good argument that the enzymes are evolutionarily related.

I also concur with what other posters have said about comparing each enzyme first with closely related sequences, for example with pfam or psiblast. Then compare the profiles derived for each local family. If, for example, the "motifs" shared between the animal and plant enzymes are also shared within all animal-derived sequences and within all plant-derived sequences you have a strong case. If the motifs you describe in your question are *not* conserved within animals and within plants then they likely represent random noise.

This is precisely what I found, for instance, when I determined the structure of the Cdc25A phosphatase:

http://www.sciencedirect.com/science/article/pii/S0092867400811903

It had been assumed that because it had a "CxxxxxR" motif, it would look like the other tyrosine phosphatases. But it shared no structural motifs with other phosphatases and in fact shared a motif with a far more ancient enzyme. At the time it required a crystal structure to reveal this relationship, but today one would be able to reach the same conclusion on the strength of careful profile analysis.