The primary output from phylogenetics programs is tree diagrams depicting the relationships among organisms.Â Very clean, very crisp, very precise diagrams.Â Precision isn't in itself a problem, but for the human foible of mistaking precision for accuracy.
I'm not interested in a precise estimate of evolutionary history so much as a correct one.Â I'm reminded as much when I see my estimates change from one precise conclusion to another as I add more data from more species.Â Â The razor-sharp output from the algorithms is seductive.
Are you using molecular data or strict character matrices? Or some sort of hybrid approach?
The tree in the figure above is strictly molecular: about 4kb of data from 5 genes. I also have character matrices for other traits, like geography and morphology, but I tend not to use those to generate the trees so much as to study trait evolution using the molecular trees as a framework.
Shifts in tree structures based on molecular data should not be that surprising. After all, we know that very different organisms can derive similar solutions to evolutionary pressures, whether they are related or not. Molecular or metabolonomic relatedness does not necessarily follow genetic relatedness.
I agree that trees can be deceiving and , as humans, we are predisposed to discern "hard targets" when processing patterns. Are you using Bayesian or bootstrap algorithms for tree development? If so, try using the confidence numbers on the branches graphically in the display. That is, something like shading or fuzzing out specific branches depending on their confidence. High confidence branches are bold and dark while less confident ones fade into the background. Could try colors too.
I've been wondering about the amount of correctness seen in molecular phylogenies versus morphological phylogenies recently. My adviser has been on my case about the phenetic nature of molecular phylogenetics and how it is a poor replacement for morphological parsimony based methods recently, and I was wondering if you would comment on that. Thanks.
I admit that whenever I hear the critique that molecular phylogenetics is "phenetic" I tune out, as that phrase shows a failure to understand what the algorithms are actually doing.
Phenetics is uses clustering algorithms. These treat data in a profoundly different manner than model-based algorithms which attempt to fit explicitly evolutionary hypotheses to data.
If there is an argument to be made via morphology vs. molecules, it isn't about the type of algorithms used- after all, similar sets of algorithms can be used on both molecular and non-molecular data. Rather, it is that the current emphasis on molecular data does little to encourage the training and employment of people to study organismal morphology. We are losing some valuable specialized expertise.
I think the argument was that phylogenies based upon overall similarity rather than the special shared characters of homology was very similar to phenetic techniques, and is therefore less likely to yield a correct phylogeny. What constitutes a robust molecular homologous character, a single base pair, a whole coding sequence, a part of a coding sequence, a series of genes? Perhaps maximum likelyhood and baysian analysis are simply a whole different and new school of systematics, equal and sister to cladistics.
I agree that molecular is a useful tool to add to the systematics toolbox. Rest assured, I am learning and employing morphology based techniques for my research, so it at least will be carried through this generation. I prefer them, but not to the exclusion of other tools. :)