The best way is to simply run a phylogeny and compare likelihood values. The alignment with the likelihood value nearest 0 is the winner. Here you could use RAxML. Otherwise you could run modeltest or protest (depending on your input) to extract both the best evolutionary model and the likelihood score!.

regards

Thanks Andrei. I am looking for a tool to evaluate the post-alignment aligned sequences to find out the best alignment tool.

The best way is to simply run a phylogeny and compare likelihood values. The alignment with the likelihood value nearest 0 is the winner. Here you could use RAxML. Otherwise you could run modeltest or protest (depending on your input) to extract both the best evolutionary model and the likelihood score!.

regards

Hi Asif

Personally I like MAFFT

http://mafft.cbrc.jp/alignment/software/

It performs particularly well when you are dealing with a huge number of sequences. You can get phylo-trees directly from it and also distance matrices. Also people involved in the development of the software are extremely competent and friendly. They have helped me several times and they are open to solve problems.

Best. Paco

Hi Asif,

There are a number of measurements, such as the sum-of-pairs score, which allow a comparison of different alignments. However, it is basically impossible to identify the best alignment tool without knowing the true alignment.

I personally prefer mafft as well, but I am aware that it is not applicable to all my problems. Sometimes Dialign-TX (http://dialign-tx.gobics.de/) or some other tool performes better.

For starters, take a look at

Thompson JD, Linard B, Lecompte O, Poch O (2011) A Comprehensive Benchmark Study of Multiple Sequence Alignment Methods: Current Challenges and Future Perspectives. PLoS ONE 6(3): e18093. doi:10.1371/journal.pone.0018093

Best, alex

For an in-house set of sequences, one idea might be to calculate to calculate, e.g. shannon entropy for each column of the alignment. You can then pick a reference sequence and plot the entropy scores for each method across all sites in the sequence. The best way to visualize this might be with a violin plot.

You have many options, so you need to know the differences what you need to do with it. So I would recommend you to read the attached paper for different MSA programs.

Article Assessment of molecular (dis)similarity: The role of multipl...

Generally I agree with Russell. The calculation of ML score is quick and efficient way how to estimate the quality of the alignment. RAxML is probably the most accurate tool for this purpose (in comparison with FastTree, GARLI, IQPNNI and MrBayes). See:

RAxML and FastTree: Comparing Two Methods for Large-Scale Maximum Likelihood Phylogeny Estimation

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0027731?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+plosone%2FComputerScience+(PLoS+ONE+Alerts%3A+Computer+Science)

RAxML-VI-HPC: maximum likelihood-based phylogenetic

analyses with thousands of taxa and mixed models (includes fair comparison with GARLI, PHYML, IQPNNI and MrBayes)

http://bioinformatics.oxfordjournals.org/content/22/21/2688.full.pdf+html

I assume your ultimate goal is a gene or species tree. Am I correct? So, if you choose your alignment based on its performance in ML analysis, what is a tree based on this alignment worth? A tree is merely a hypothesis derived from the underlying data. If you are selecting the alignment based on your expected outcome in tree reconstruction you are manipulating the data.

Hi Asif,

you can use NORMD for this purpose. I often use AQUA which actually performs alignments using different alignment methods and choses the best: http://bioinformatics.oxfordjournals.org/content/26/2/263.full

hope this helps.