Hi Felix,

There are several possible causes (i.e. homoplasy, a recombinant strain, few informative sites, etc).

Does the tree without your sequences have good statistical support (i.e. Bootstrap)? 

How many informative sites have your alignments?

Are the phylogenies of different genes congruent?

If you include your sequences, has the concatenated tree less support?, have phylogenies of different loci less congruence?

Answering these questions should help to answer which is the cause of reordering. I can help you to understand why this is happening but I need more information.

a phylogenetic tree is not an absolute answer but a relative one ; relative to the sequences used, relative to the method used ; try another outgroup? some sequences in databank are not correctly named (taxonomically), some sequences are not correct at all!

Hi Félix,

I have a few questions, what are the statistical  values for the nodes related with the observed changes? are they significant? Do you employed the same parameters, such as nucleotide substitution model, for both trees? When you said "large clusters" I assume diferences in basal nodes, and probably you expected the including of your sequences in the main topology of the tree observed in the article, is this true?

 Hi Felix,

Cluster reordering by including additional strains may be caused by several causes (i.e. homoplasy, recombinant strains, few informative sites, uncertain alignments).

There is few information in the question to know the cause. However, some questions will help you to answer it.

First, did you check the alignments? are they right? (i.e. no gaps, no Ns)

-have the alignments enough polymorphic and informative sites? Phylogenies based on sequences with few information are susceptible to reordering if the newly added sequences had homoplasy.

-Is the phylogeny without your sequences well supported? (i.e. bootstrap?) if a phylogeny is not strongly supported (particularly the cluster that you are interested in) may be affected by the addition of new strains.

-Are the phylogenies of different loci (without your sequences) congruent among them? If there is high incongruence, clusters may be simple artifacts of concatenation.

-If  the answers to previous questions are "yes to all" and consequently the published clusters are well supported, you should look if congruence and support are affected when you add the your sequences. If support and congruence is affected, you may infer that your sequences are probably recombinant. You should use other methods to confirm it.

this article may help you. It describes  a software for MLST analysis in diploids but it works with bacteria too. The software was designed to analyse problems similar to this.

http://www.ncbi.nlm.nih.gov/pubmed/24025589

Hi Felix, 

I have no much idea about the issue you are facing, but you can look into few basic things prior to phylogenetic construction

1) Set of sequences you are aligning, are they highly distant phylogenetically (sequences belong to different phylum or order or family)? if so, then you can try other alignment methods (MUSCLE) which does a combination of both global and local alignment  

2) What is the length of your sequence? is there variation in the length of your sequence and sequences from the article? if so. then you probably need to trimmed sequences after MSA, making these set of sequences to same length and then go for phylogenetic tree construction. try this with both alignment method (CLUSTAL W and MUSCLE).

Hi,

In my experience, this is a fairly common problem when you are are running a ML analysis using a molecular marker with a fast rate of evolution, and including a complex mixture of both recent and deep divergences. In that case, the topology of small clades will be quite consistently resolved, but the relationship among these small clades can strongly vary even with very subtle changes to either alignement, model of evolution and/or taxonomic composition of the data matrix. This is due to the fast rate of evolution overprinting the evidence for deep-time branching pattern, and will lead to a seemingly random placement of big clades. Maybe this is what's happening to your dataset. I hope it helps!

First, thanks to all.

Information detail:

I think that the metodology is correct, because when i repeat the phylogenetic tree, using only data from the article, i obtained his same result. The only changed that i consider significative is the length, as Shaikh Mohammed Inayat mention. In the article the concatenated lenght is 3100pb and my concatenated is of 2700 pb. So i did a proof. Using the same methodology, i construct the phylogenetic tree, using only data from the article, but shortening until 2700 pb. My surprise is that appear the same problem, a topology change of the same group. So i think that the concatenated lenght is the principal factor of this topology change in the tree; and it is not due to my sequence.

Changes of parameter α, change quantifiable relation between the richness of a monophyletic groups and its relative clade.