This was the original question. Dear Evoldirs, Recently, I am trying to identify the best estimated tree using 10 nuclear DNA intronic sequences.  As you might know, each nuclear gene might have different evolutionary rate.  I've seen many papers dealing constructing tree by combined nuclear DNA dataset.  My question is if it is the best to estimate the tree using many nuclear DNA markers which has different evolutionary rate.  How about is the case for mitochondrial genome? Some use several mtDNA fragments and other use complete mtDNA. In mt genome, 13 coding genes, several rRNA, tRNA, and control region are present. If you use only 2~3 selected mtDNA markers to draw tree, would it tell you the same tree with the one drawn from complete mtDNA? Are there different evolutionary rate for each genes of mtDNA as well? How different are they?  Are there good references to read? Please let me know. Best Junghwa AN 1) Hello Jungwha, Mt genes can have different evolutionary rates - we found this out with some of the work we did (papers attached), and Jeff Palmer has found this as well. I'd be curious to hear what kind of responses you get - would you post them on evoldir when they come in? Thanks! Camille 2) Hello Jungwha, Well, I will start off by saying that I'm not a phylogeneticist. But I was just talking to one last week about using multiple loci for making trees, and he was making it very clear that all genes have independent evolutionary histories, so simply concatenating a handful of genes doesn't necessarily give you a more complete history of the evolutionary history of that organism. I think the way to deal with that is to publish the individual gene trees along with any concatenated tree. In that light, it shouldn't necessarily matter that much if some of the loci are linked, but you could fairly easily test that presumption by testing for linkage disequilibrium among your loci. As for paralogy, I am not sure of all the ways to test for that, but I do know that if you translate any of the coding portions of the genes you sequenced and find stop codons in the middle of the gene or perhaps high dN/dS ratios, that can indicate a lack of functionality and thus perhaps paralogy. If you are referring to nuclear paralogs of mitochondrial genes, that certainly happens, and there are some known cases of that. Because mutation rates are so different between nuclear and mt genes, you could take a look at your substitution rates and see if they are dramatically lower in any of your suspected paralogs - that could indicate nuclear paralogs. Camille 3) Hi, as each dataset is different, you may try at first to do separate analyses for all regions and compare the trees. If they are incongruent to a large degree, it is not a good idea to combine the data. Best, Judith Dr. Judith Fehrer Institute of Botany Czech Academy of Sciences Zamek 1 25243 Pruhonice Czech Republic 4) I can't say I've done this, so I'm sure you'll get more experienced responses, but I doubt that you would see as large of differences with coding regions in mitochondrial DNA as you would in nuclear DNA.  The main reason is that mtDNA should really be considered one large linked set of genes (due to size and how quickly it evolves).  Using multiple mtDNA coding regions is more like pseudoreplication in statistics than actual replication, whereas if you used several unlinked coding regions of nuclear DNA, you should be getting actual replication. Julie 5) As to control for mutational variance, long sequence might help, e.g. long segment of mtDNA. However, it only tells you one particular gene history (No recombination inside). However, 10 independent nuclear DNA intronic sequences have different coalesent histories. so best way to do it is not combine them together. Rather, you should use ways that recently burgeoning in this field, like programs BEST, Bucky, things like that. suggested readings: Maddison&Knowles 2006, Syst. Biol. 55(1):21-30 Edwars, Liu & Perl, 2007, PNAS, 104:5936-5941 Kubatko&Degnan, 2007, Syst.Biol.56(1):17-24 Hope this helps you. Best, Qixin He 6) I good place to start is reading papers by S. V Edwards and his group. A recent one is Edwards S, V. (2009) Is a new and general theory of molecular systematics emerging? Evolution 63, 1-19. They also have a good program to use for multiple genes called BEST. This program gets around problems of simply concatenating the sequences with different rates. Good luck, Alicia Toon 7) In this case, I'm referring to statistical replication rather than cell or DNA replication.  When you are trying to ascertain relationships between species, the reason we want many characters is because we need multiple types of evidence that reflect the same evolutionary pattern.  This is why it is so important to use unlinked genes: assuming there is no selection on them and they recombine freely, they should evolve independently.  Therefore, if they all give the same relationship between taxa, this is good evidence that the relationship is true. On the otherhand, if you use genes that are linked to one another, they should all show the same evolutionary pattern not just because they have the same history, but because any forces on one of the genes (such as selection) occurs, all of the others will be dragged along.  Therefore, linked genes really function like one single gene because they don't evolve independently.  Maybe this will help: http://en.wikipedia.org/wiki/Pseudoreplication I hope that helps.  It is hard to describe in an e-mail. Julie 8) if those genes are on different chromosomes, they are definitely unlinked genes!! as for paralogs, you should be more careful...But I'm not so familiar with how to check for paralogs.   Qixin 9) Unless we have a genome or linkage map to work with, we have to assume unlinked loci.  Often, you can test for linkage using Hardy-Weinberg tests.  (I can't remember what sequence analysis programs do that.)  I think choosing your sequences based on a good reference species is a good way to go (in terms of trying to make sure your genes are unlinked).  I wouldn't be able to tell you if your genes are unlinked or not, so if I were you, I would look to see if any of the software you are planning to use to analyze your data checks for linkage disequilibrium or not.  (I know some of the haplotype software does.)  As for the paralog gene, I don't have any experience identifying paralogs.  However, I saw several good websites come up when I Googled identify paralog, so I would try that and see where it takes you.  I know there must be ways to check (maybe using cloning combined with a McDonald-Kreitman test), but I'm not sure.  Good luck! Julie An Junghwa