Convergent Evolution - Reasonable conclusion? or convenient excuse?

Hi Bob,This is interesting research, but I don't think it necessarily supports your contention. Actually this isn't what it says, it says that the same amino acid substitutions seem to have ocurred. I'd suggest that what their paper (Liu et al., 2010) should talk about is Prestin protein having undergone sequence convergence since the gene sequence more normally refers to the DNA sequence.So the same amino acid substitutions are distinct from the same mutations. The same amino acid substitution can arise from a number of distinct mutations depending on the amino acid in question. There are multiple possible nucleotide sustitutions giving rise to the same amino acid substitutions so the chances of convergent amino acid substitution are considerably greater than those of convergent DNA nucleotide substitutions.In terms of the DNA sequences the paper specifically states ... ... so in fact the gene sequences, the level at which the mutations actually occur, don't show convergence, but the protein sequences, the higher level phenotype, do.We are talking here about 14 amino acid substitutions and it is worth noting that value is a collective one between all the dolphins and all the bats. In other words there is no single bat-dolpin pair you could pick that would show those 14 convergentamino acid changes.So I would echo Dr. A's suggestion that if you really feel that this is still nearly impossible you should show your workings. And remember that the basis of that calculation should not be 'The statistical odds of the exact same mutations being selected in both species' but rather the statistical odds off the same amino acid substitutions occuring in such a pattern that between 14 species of bat and 4 dolphin species they can find 14 well supported convergent sites.It is also worth noting that when the authors include the sperm whale in their analysis your whole premise falls apart. Suddenly the dolphins end up back in their usual place in the phylogenetic tree. So in fact this research still supports the Darwinian model and indeed seems to fit in with the traditional nested hierarchy. The only way it doesn't is if you leave out some of the data. The thing is this simply isn't true; the sequence similarities between human and chimpanzee are orders of magnitude stronger than the ones shown in the paper stretching across the whole genome and are at the more relevant actual genetic sequence level.TTFN,WKEdited by Wounded King, : used wrong term swapped conserved for convergent.Edited by Wounded King, : Added link to abstract.

Looking again at that issue of Current Biology I see that there were in fact 2 papers on the same topic. The other paper (Li et al, 2010) has a less extensive analysis but shows essentially the same pattern.The one particularly distinct analysis they do is to instead of just looking at the whole nucleotide sequence for the gene look at specific subsets of nucleotides, those leading to synonymous and non-synonymous substitutions. They show that looking at the non-synonymous substitution sites alone also produces an anomalous tree position.I don't see that this has much bearing on my previous comments though in terms of genetic support for the nested hierarchy.TTFN,WK

I'm not sure how that is another element to consider, since it is the whole point of both articles. They both posit convergent protein sequence evolution of Prestin as a result of selective pressures associated with echolocation.But as I have been pointing out, protein sequences and amino acid substitutions are not really genetic data in and of themselves. Selection for the resonance response only selects for the same amino acid sequences not the same DNA sequences. While this selection will affect the DNA sequences to some extent it still allows a lot more variability than at the amino acid level.For Bob's argument to make any sense the nested hierarchy really needs to fall apart at the genetic level, it is no good him just choosing an arbitrary phenotypic level and then saying, look at these supposedly convergent traits they blow a major hole in the darwinian nested hierarchy approach.TTFN,WKEdited by Wounded King, : Fixed ridiculous spelling error

These sort of calculations are notoriously unreliable and massively sensitive to initial assumptions.For a start, as I mentioned previously, it isn't the case that any one dolphin-bat pair actually shows all 14 convergent amino acid substitutions. Almost certainly not. If we assume that there is an original ancestral Prestin sequence then we need to calculate the odds of the specific changes from that sequence, not the odds of a particular set from of a random assortment of 14 selections from a pool of 20 possible amino acids.Some amino acid changes are simpler than others and because nucleotide substitutions are not completely random, in that certain changes are more frequent than others, certain amino acid substitutions will also be more frequent and therefore more likely to occur. I'm not really sure why you feel these figures are particularly relevant. For a start we don't even have a per site mutation rate for the Prestin gene, so the proportion that will be deleterious is somewhat irrelevant. I'm also a bit dubious of the 70% figure, the vast majority of mutations are neutral, in a coding gene this is different but I still find the 70% figure unlikely. The first thing I found sounding similar to this was the wikipedia article on mutation which gave 70% as the figure for deleterious mutations from mutations causing non-synonymous amino acid substitutions.I'm also not sure why you want to treat deleterious mutations as distinct from null or disabling mutations.As I said these sort of calculations are essentially pointless.TTFN,WK

You seem to have already forgotten the point we discussed earlier about the appropriate level for convergence that was relevant to casting doubt on modern evolutionary theory being the genetic level. Almost all of the examples in that list are of convergent morphology, which is by no means the same as convergent genetics. If you feel you can make a case that morphological convergence between mammals and marsupials, for example, is due to genetic convergence then feel free. Many of these cases directly contradict your assumption that convergent morphologies rely on convergent genes. Convergent morphologies or functions can arise from totally distinct genetic sequences or complements.There are only 7 examples for enzymes or biosynthetic pathways and even in those cases there is no evidence of genetic convergence.TTFN,WK

Can you think of anything that would? And does it make a more reasonable conclusion than the one we already have which is supported by the material processes we can observe and understand?TTFN,WK

Man! Professor Branestawm, that takes me back.TTFN,WK

The second list cannot be placed in a nested hierarchy, no matter how hard you try.

I'm with Mr. Jack on this one, I don't see why you couldn't put those into a nested hierarchy. I was planning to run your examples through some of the PHYLIP programs myself but Mr. Jack saved me the bother.That said I think it would be interesting to see what happens if Jack runs those two sets of data through the seqboot program first to produce a bootstrap analysis. The real distinction between genuine sequences and randomly generated ones is not tha ability to order them in a nested hierarchy but the ability to do so in such a way that they are significantly distinct from the results one would get from ordering random sequences in a nested hierarchy.So I would predict that for a set of genuine homolgous DNA sequences against a set of randomly generated sequences you would get a well supported tree for the genuine sequences and a very poorly supported tree for the random ones. As far as your examples go, I'm not sure how those would fare in such an analysis since they need to be analysed as characteristics rather than as sequence data. I think the tree length distances Mr. Jack notes are also an important distinction.TTFN,WKEdited by Wounded King, : misspelling of fare

Hi Percy I just ran the program using the input file ... And got the same tree as Mr. Jack.

+--------AAAEGBA
           +--------6
           |        |        +AAAEAEA
           |        +--------5
           |                 |        +----------------AAEGAAA
  +--------3                 +--------4
  |        |                          +AAAGAEA
  |        |
  |        |        +--------GAEAABA
  |        +--------2
  |                 +--------GAAGABA
  |
  1--------AAAAAGC
  |
  +--------AFAAAGA

You can get the Pars program along with the rest of the PHYLIP package at the link in Message 82.I tried bootstrapping that first set of data and got this ...

+-------AFAAAGA
                          +--64.0-|
                          |       +-------AAAAAGC
                  +--19.6-|
                  |       |       +-------GAAGABA
                  |       +--42.4-|
          +--32.6-|               +-------GAEAABA
          |       |
          |       |               +-------AAAGAEA
  +-------|       +----------20.7-|
  |       |                       +-------AAEGAAA
  |       |
  |       +-------------------------------AAAEAEA
  |
  +---------------------------------------AAAEGBA

Where the branch numbers represent how many time out of 100 those particular groupings came out in the bootstrapping.Sadly for your argument Percy the bootstrapping of the random data sets actually gives an arguably more robust hierarchy in that most of the branches are better supported.

+-------ADAGGDG
                                          +--80.7-|
                                  +--46.6-|       +-------FAAGGDA
                                  |       |
                          +--56.9-|       +---------------FDDCEBG
                          |       |
                  +--28.1-|       +-----------------------FBFGGBF
                  |       |
          +--50.6-|       +-------------------------------ECEEBBB
          |       |
  +-------|       +---------------------------------------EBDCCCE
  |       |
  |       +-----------------------------------------------EGGAAEE
  |
  +-------------------------------------------------------CFCGAEE

I think what this really shows is that 7 characteristics is a very small sample to draw any sort of conclusion from.TTFN,WKEdited by Wounded King, : Added bootstrapped tree.Edited by Wounded King, : Added bootstrapped tree of random data sets.Edited by Wounded King, : Because I can't tell the difference between 7 and 8, that'll teach me to try arithmetic without using my fingers.

No, because the length is cumulative for all branches so I make Percy's tree out as having a length of 14. Also Percy's tree has a quite different topology because he has an ancestral sequence making it a rooted tree. The Pars program also doesn't correct for multiple changes overlaying each other, so the 2 changes at position 4 will confound it.Adding in the ancestral sequence as an outgroup still doesn't give me Percy's tree exactly, but it does produce a tree with a length of 14.

+AAAAAAA
  |
  3-----------------------AAEGAAA
  |
  |           +-----------------------AAAGAEA
  |           |
  |           |                       +-----------AAAEGBA
  |           |           +-----------5
  |           |           |           +-----------AAAEAEA
  +-----------1-----------4
              |           |           +-----------GAEAABA
              |           +-----------2
              |                       +-----------GAAGABA
              |
              +-----------AAAAAGC
              |
              +-----------AFAAAGA   

Another big difference is that the Pars program cannot be contrained with Percy's 1 step mutation condition.Running it with slightly diffferent variables the program actually produces an alternative rooted tree with a tree length of only 13.

+AAAAAAA
  |
  4-----------------------AAEGAAA
  |
  |                       +-----------AAAEGBA
  |           +-----------6
  |           |           |           +AAAEAEA
  |           |           +-----------5
  |           |                       +-----------AAAGAEA
  |           |
  |           |           +-----------GAEAABA
  +-----------2-----------3
              |           +-----------GAAGABA
              |
              |           +-----------AAAAAGC
              +-----------1
                          +-----------AFAAAGA 

The program also produces a table which lets us see what ancestral states it has inferred.

From    To     Any Steps?    State at upper node
                             ( . means same as in the node below it on tree)

root      4                AAAAAAA
   4   AAAAAAA      no     .......
   4   AAEGAAA      yes    ..EG...
   4      2         yes    .....B.
   2      6         yes    ...E...
   6   AAAEGBA      yes    ....G..
   6      5         yes    .....E.
   5   AAAEAEA      no     .......
   5   AAAGAEA      yes    ...G...
   2      3         yes    G......
   3   GAEAABA      yes    ..E....
   3   GAAGABA      yes    ...G...
   2      1         yes    .....G.
   1   AAAAAGC      yes    ......C
   1   AFAAAGA      yes    .F.....

TTFN,WKEdited by Wounded King, : No reason given.

I can't really follow your reasoning here, Taq wasn't saying that he could identify any specific date that a particular mutation ocurred, simply that there is no reason to presuppose that all of the mutations needed to occur simultaneously in the same organism. What can be done is to identify the most parsimonious model for the diversification of related sequences from an ancestral sequence.The way deletions are identified in genomic analysis, when they do occur, is by comparison to reference sequences. These will usually be taken from a consensus sequence from a sample of the population or in cross species comparisons from a putative conserved ancestral sequence.In most cases we also have to infer the ancestral sequences from extant sequences, although now there are some long term data sets, the Lenski long term evolution experiment one in particular, where we do know exact ancestral sequences. That isn't what happens, as Jar suggests the mutations are alterations of existing genes, not the deletion and wholesale replacement of one gene by a totally novel gene.TTFN,WK

But we aren't discussing the de novo creation of the entire Prestin gene, we are talking about a small number of amino acids that are common to the bats' and dolphins' Prestin protein. The Prestin gene was already extant in the mammals, it didn't suddenly appear in dolphins and bats while being absent in every other mammal .What topic exactly is it that you are trying to address?TTFN,WKEdited by Wounded King, : No reason given.