I don't know which kind of molecular data you want to align but I would advise the use of Feature Frequency Profile Phylogenetics Package. FFP is a class of alignment-free methods suitable for the whole genome comparisons. You can use it either to construct an FFP profile from nucleic acid sequences or from amino acid sequences. The software also include other utilities such as distances/similarity metrics and bootstrapping or jacknifing permutation.
the package is available at: http://sourceforge.net/projects/ffp-phylogeny/files/
You might want to consider an approach that treats the alignment as a model parameter in addition to topology, branch lengths, and substitution. There is a very good Bayesian estimator executed using the software Bali-Phy. See the following concerning the theory and implementation:
Redelings BD and Suchard MA Joint Bayesian Estimation of Alignment and Phylogeny, Systematic Biology, 54(3):401-418, 2005
Suchard MA and Redelings BD BAli-Phy: simultaneous Bayesian inference of alignment and phylogeny, Bioinformatics, 22:2047-2048, 2006
There have probably been more recent tools developed, but I think Bali-Phy has been well explored and has been used in a good number of research articles to give you some point of reference.
Three of the four earlier answers dealt with how to align data for phylogenetic analysis, not how to do phylogenetic analysis without alignment.
In response to the question asked, I suggest using kSNP (Gardner, S. N. and B. G. Hall. 2013 When whole-genome alignments just won't work: kSNP3 v2 software for alignment-free SNP discovery and phylogenetics of hundreds of microbial genomes. PLoS One 8(12): e81760. doi:10.1371/journal.pone.0081760).
kSNP is a SNP identification program that does not require genome alignments or use of a reference genome. It identifies SNPs, and estimates phylogenies from those SNPs by parsimony, Neighbor-Joining and Maximum likelihood methods.
kSNP is freely available as executables for Mac OS X and Linux at https://sourceforge.net/projects/ksnp/. The current version is kSNP v2, but kSNP3 is ready to be released as soon as officially approved by the Lawrence Livermore National Laboratory. kSNP3 is significantly improved, but input parameters are different from kSNP v2. I suggest waiting 2-3 weeks until kSNP3 is available on SourceForge
MEGA is not an alignment-free method. It is a program that can align sequences and estimate phylogenies from those alignments, but phylogenetic analysis with MEGA absolutely requires alignment.
i have gene sequences trying to align free phylogeny but how to select proper method? In the literature many methods are described for alignment free phylogeny such as Feature frequency profile (FFP), Composition vector (CV), Return time distribution (RTD), Frequency chaos game representation (FCGR) and Spaced-word frequencies. Which is best one for gene sequences alignment free phylogeny??
Definitely you want to use POY, as suggested by Santiago. This provided that your sequences are HOMOLOGOUS, of course. POY can infer both parsimony and maximum likelihood trees.
Dynamic homology (the approach implemented in POY) is a highly complex problem, which traduces in that your analyses will take long to complete (several hours or days; the larger the number of sequences, the longer the computational effort). This can turn the analysis prohibitive even for moderate size datasets.
You might want to take a look at this paper: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3299439/
and also: Varón, A., L. S. Vinh, W. C. Wheeler. 2010. POY version 4: phylogenetic analysis using dynamic homologies. Cladistics, 26:72-85, and citations therein.
I think people have discussed a few tools that fall broadly into alignment-free phylogenic estimation methods (FFP, kSNP, POY, and Bali-phy that Barry pointed out is not technically alignment-free but integrates over uncertainty in alignment along with uncertainty in other model parameters). Something that has not been discussed is what are your goals, data, and why you require alignment free phylogenetic estimation. The method should be appropriate for your data and your question, as the motivation and assumptions for these methods are quite different.