Felipe is right about considering what story you are telling with the protein in question. If you are focused on just a single functional domain then you will only be able to infer it as that single functional domain.

We have used DnaSP, a free software (http://www.ub.edu/dnasp/). If you are looking for single amino acid changes in highly conserved functional domains, it will give you a clear substitution rate (dN/dS). This will indicate the "strength" of the evolutionary forces/process.

hi Manan,

Your interpretation should be restricted to the domain that you are studying. The evolutionary forces on a conserved domain vs an unstructured region would be quite different, so their evolutionary histories. Nevertheless the domain analysis could be quite informative.

Hope this helps,

Diego

I think the term "evolutionary forces" can be a bit misleading. The underlying evolutionary processes are exactly the same. What changes is the exact physical restrictions for the function which nevertheless would point to the same conservation of those regions important for the function.

Anyhow, i've seen quite a lot of times that modular proteins of that sort usually are a composite of several unrelated domains gained in the course of time. For instance, in the case of transcription factors you would see that different subfamilies under the same family of proteins have completely non-homologous domains. In that case splitting into pieces is the only way of doing the analysis.

Just keep in mind, as it was mentioned by Diego, that your analysis is only valid for the domain in question and not necessarily for the story of the complete protein,

yes, perhaps I should have qualified these words. It should be the "strength of the evolutionary forces/proceses" would be different between these different regions. One could be neutral and the other under different types of selection, for instance

Yes! I would recommend the use of TreeSAAP, which examines the evolutionary changes in amino acids in the context of 500+ physiochemical parameters. Here's the link to TreeSAAP and a video tutorial (http://dna.cs.byu.edu/treesaap/). First thing you need to do is build a good quality input tree (MrBayes, or something like that). We have published a paper outlining an scheme that uses many convergent approaches, one of which was TreeSAAP, to answer the evolutionary changes in specific domains.

Article Functional Validation of Hydrophobic Adaptation to Physiolog...

Felipe is right about considering what story you are telling with the protein in question. If you are focused on just a single functional domain then you will only be able to infer it as that single functional domain.

We have used DnaSP, a free software (http://www.ub.edu/dnasp/). If you are looking for single amino acid changes in highly conserved functional domains, it will give you a clear substitution rate (dN/dS). This will indicate the "strength" of the evolutionary forces/process.

I'm assuming that you are interested in the C-terminal domain because you want to or have identified the active sites. In this case, the fundamental evolutionary unit is not the entire domain, but the codon. For investigating selection on codons, you'll want to use a Branch-Site method, implemented in both HyPhy or PAML. Also, these methods are fairly robust to phylogenetic uncertainty and effective whether your comparisons are at the population or species level.

I'm not familiar with TreeSAAP, but you'll definitely want to keep functional properties in mind and how this might affect sidechain affinity.

Please, let me answer with a question. What kind of evolutive analyses do you want to perform taking into account the protein structure?

In my opinion, the first approach you have to do in an evolutive analysis is over the protein sequence variation. To this aim, you need to collect sequences from different species, perform a sequence alignment and analyze this data using the afore mentioned software (PAML, HyPhy, PHYLIP, et cetera). Once you've got the evolutive description (substitutions rates, shannon entropy, polimorfisms, etc.), you could look for correlations between this results and any structural variable. Thus, if you are limited regarding the availability of your structural data, as Dr. Riaño-Pachón said: "Your interpretation should be restricted to the domain that you are studying". I am strongly agree with him.

I invite you to have a look at this paper. It is a simple study where we studied the differential evolutionary dynamic between different regions (surface and interior) of two proteins.

https://www.researchgate.net/publication/221858765_Thermodynamic_stability_explains_the_differential_evolutionary_dynamics_of_cytochrome_b_and_COX_I_in_mammals

Article Thermodynamic Stability Explains the Differential Evolutiona...

Thank you all for your suggestions.. they are quite helpful indeed...

As pointed out by Diego, my interpretation will be restricted to the (C terminal Domain) domain and not the entire protein, which is fine. As Felipe has given an example of transcription factors, my protein also belongs to one such family with several other non-homologous members. As a result i sought out to dissect the protein and consider studying the evolution of the common/conserved domain of the entire family. Also thank you for the tools i shall go through the tutorials and use them

Hi again, I did try generating a tree based on Cterminal region alone, but the resulting tree was very weak .. with

Where is your data coming from? Are these translated nucleotide sequences or amino acid sequences obtained through HPLC-MSMS? Also, what is the scope of the study (single population, multiple populations, multiple species, multiple individuals per species), and number of individuals you are observing? Also, are you trying to estimate the phylogeny from a single ortholog or the entire gene/protein family?

If you have nucleotide sequences, translate them to amino acids and align with MUSCLE (which you have already done), but then map the aligned amino acids back to the codons. I suggest this because most of the information is probably synonymous mutations. If you are not comfortable coding this mapping process in python or perl, you should use TranslatorX to accomplish this task. If you only have one alignment to perform, TranslatorX has a webserver you can use. Just make sure you check the alignment afterwords, because TranslatorX will guess the reading frame; depending on the quality if the data, presence of framshifts, or phylogenetic distance between taxa, you can get some poor results.

If you only have the amino acid sequence, still use MUSCLE. The only case where MUSCLE will not be your friend is an exceptionally large dataset covering a large phylogenetic distance. If this is your scenario, try using Sate. This is a tree-guided alignment algorithm that accounts for inheritance.

No matter the form of your data, you may want to try using RAxML for estimating your phylogeny by generating a reasonable number of standard bootstrap replicates for your dataset. Then you can make a bootstrap consensus tree and reoptimize the branch lengths.

As a side note, trees with low support or polytomies are not necessarily a bad thing, especially in studies of molecular evolution. For population-level work, a neighbor-joining or UPGMA tree (You can make both of these in MEGA) is usually sufficient. In a single population, you might test if your phylogeny does not reject a molecular clock hypothesis. If it does not, then the UPGMA tree will be correct. If you are investigating this protein region in multiple species, then you'll probably reject a molecular clock hypothesis and want to use RAxML, so as to account for rate heterogeneity with the model appropriate for the size of you dataset.

MEGA can definitely get the job done if you're investigating a single ortholog or a small dataset, but you may want to get away from it if you have access to a UNIX-based operating system. I always found all of the mega session files it generated to be bothersome and the output files would have strange line breaks or something that you would have to get rid of before using the output for something else. Also using programs outside of MEGA, such as MUSCLE give you access to all of the options which may be needed at some point in time.

I hope this helps! If you have more questions, more information about the size, scope, and type of information for the dataset would help produce a more refined answer. I focused on the least computationally expensive options here.

Hello George,

Thank you for the detailed answer..it is quite helpful. The sequences i am using are amino acid sequences obtained from UniProt database. The sequences are of a single protein and its homolog found in different species, so its a heterogenic population. One of the suggestions you have given is Sate. How reliable is Sate? Also when using the RaxML option in Sate, the resulting tree does not show any bootstrap values at the nodes. So another websever providing RaxML would be better?

Hi Manan,

Sate is reliable only when you have greater than 50 sequences. The developers stress that this is a tool for large and complex alignments. If there is too little data, the likelihoods from the tree decomposition cannot be discerned, hence some set of alignments will be equally likely.

The RAxML tree from Sate is only meant to help guide the alignment algorithm, so as you have noticed, you'll have to generate your publishable tree some other way. Luckily, RAxML has many implementations to suite whatever your computational needs may be. If you would like to download and install the software to run it locally, there are both command-line versions and (I think) a GUI application available. If time or processor availability is an issue for you, you can use the CIPRES data server. I've personally never used it, but it has been a tremendous help to many people and I think all you need to do create a login identification and give them a little bit of information about where you are at. There seems to be strong documentation and many user groups for utilizing the CIPRES resources, so it shouldn't be too hard to get started.

Also, if you're final result is an unresolved phylogeny and there is variation present in your protein data, you can always estimate a species tree from some other source of data and reoptimize branch lengths from the protein data. Just a thought, but hopefully everything falls into place on the first pass.

Thank you again George.