It depends on what is the total number of sequences, can you give an idea?

For more than 300 - 500 sequences !

When you have that many potential sequences, you might want to consider what exactly you are wishing to convey with your phylogenetic tree.  For example, would it perhaps be better to have more than one tree where each tree is made up of more similar sequences?

I would suggest using a simple analysis with a neighbor-joining approach (MEGA program), and from that break your sequences into smaller groups on which you can then run more rigorous tree-building algorithms like Bayesian inference (MrBayes program).

I did phylogenetic analysis with concatenated 900 DNA sequences with RaxML, but this is not very correct (every locus can have it's own evolutionary history, which does not necessary match each other). If you have banana breeds and lines which belong to same species or closely related species and can not exclude possibility of introgression, you need to use approach which account for hybridization events, there are number of softwares like TreeMix, PhyloNet, SNAPP. The last one works well with few samples, but analysis of more then dozen of samples is not feasible and you also need to extract SNPs from your sequences (one per locus). Hope, you have DNA sequences too, not only proteins... So, you see, hard to give simple answer.

Mega 6, BioEdit

·         For multiple alignments - (automatically) MAFFT v. 7.036 (http://mafft.cbrc.jp/alignment/server/)

·         Manual adjustments for improvement - BioEdit v. 7.2 and ClustalX

·         Parsimony analysis - PAUP (Phylogenetic Analysis Using Parsimony) v. 4.0b10 

·         MODELTEST v. 3.7  to find out the best-fit model of evolution for each data set for Bayesian and Maximum Likelihood analyses

·         Maximum-likelihood (ML) - performed in RAxML implemented in raxmlGUI v.0.9b2

·         Bayesian analysis - MrBayes v. 3.1.2

First of all for phylogenetic analysis you need a good alignment. I recommend tools for multiple alignments on EBI webcite  http://www.ebi.ac.uk/Tools/msa/

You can choose an algoritm  best for you type of data.  On the webcite you can also construct a phylogeny by distance-based methods.

MrBayes recommended above is very powerful tool for phylogeny, but I am not shure that it can work with aminoacid sequences. MrBayes  can analyze DNA, restriction and coding morphology datasets.

For ML,  Garli (Genetic Algorithm for Rapid Likelihood Inference ) would surely suite your needs. (https://code.google.com/p/garli/ )

Try our MetaPIGA software: http://www.metapiga.org.

Best, M.