hi,

Article Patterns of Positive Selection in Seven Ant Genomes

Please see " Mitochondrial Gene Families Data Set" part of methods to find detailed procedures, data and software for molecular evolution.

Best,

You could check Genbank to see if sequences already exist for the taxa you are interested in, download them, align them in Seaview using Muscle (or online using Muscle) and run them through Phyml on Seaview with a number of different models. Check the likelihood scores of each model and choose the one with the higher log likelihood (lower negative log likelihood) for your final tree. I am guessing that, on the basis or my own work with insect trees that you will be using an HKY+I+G model ore perhaps a GTR+I+G model. But do try simpler ones too.

If you are doing this right, you will use PAUP's implantation of the decision theoretic model test to determine your best model from your sequence data. This will involve making a quick neighbor-joining tree for the automatic model test to work with. Then go back to Phyml and choose the model that most closely resembles the PAUP results.

Then I assume you are using protein coding genes, so also put your data into Raxml online and specify in a separate script (a data file) tor analyse the there codon positions separately (different model parameters for each position) and be sure to link them so you get one tree that combines the results of all three positions.This should give you an even higher likelihood score that the unpartitioned analysis. So ideally you would have a good tree to map you selection analysis on.

Now go to Data Monkey and use the Genbank created tree for an independent phylogeny and enter the sequence data for the genes that you want to study for adaptive evolution. You will see a number of different tests so read each description to decide which you want and run the analysis. Be sure that the taxa names are identical in the reference tree and the sequence you analyse.

Madhav, That's just fine. But there is so much more to testing for selection than simple dn/ds. Of course that is exactly where to start and I usually do a dn/ds with Seaview where I also do the alignment and maximum likelihood.