I know of a couple of measures of signal:noise ration in phylogenetic analyses, such as the consistency index and the retention index, but most phylogenetic reconstruction software packages do not compute these values and include them with the tree output. Bootstrap values can be somewhat related to phylogenetic signal, but more often are related to the ratio of informative to noninformative (invariant within a clade) sites in the data set.
With different data sets, I would expect that the consistency index or retention index value for good data would be different. For example, with perfect sequencing (no sequencing errors at all) of human, chimpanzee, gorilla and orangutan I would expect CI values close to 1. But for perfect sequencing of of the same DNA region from mammal, frog, salamander, fish, insect, the best value expected might be closer to .6 or so. And it might be different if we have 10 widely diverse samples from each organism than if we have just one of each (10 mammals, 10 frogs etc).
For most phylogenetic discussions I see here in ResearchGate and many other places, the focus is on the models and methods of analysis. But in my experience, troubles with phylogenetic reconstruction giving impossible results is essentially always due to problems in the data, rather than problems in the model or method. Good data will create a reasonable tree with any method, but it is frighteningly easy to come up with a bad data set for example due to misannotation of genes or organisms in GenBank. And even with "perfect data" there can be problems with saturation of variable sites if we for example compare sequences from all lentiviruses, or DNA polymerase from bacteria, plants, animals, fungi.
So, if you know of good review articles covering this topic, or you know of software that is useful for checking multiple sequence alignments for various types of quality, please post them here in the answers.
Please do not forget that the consistency index (CI) measures the amount of homoplasy in a cladogram, however this measure only makes sense in a Parsimony analysis framework. Remember that in phylogenetic analysis based on explicit models (ML and Bayes) all sites are important to calculate the likelihood given the model of molecular evolution, and then the CI do no sense at all in these methods.
I like the DAMBE function (under the graphics menu) for plotting transitions and transversions vs pairwise distances, to check a data set for saturation of silent sites. Transitions (purine purine, pyrimidine pyrimidine; G A and C T) outnumber transversions because of the biochemical properties of DNA damage repair and replication error. But there are twice as many possible transversions ( G C, G T, A C, A T) so saturated sequences in theory could have up to 2 fold more transversions than transistions. I am attaching a couple of plots which illustrate this.
Given that the diversification of some taxa is very ancient, the selected molecular markers could be saturated (non phylogenetic signal), and provide spurious phylogeny. Therefore I suggest evaluating whether the sequences were saturated and thus useful for the phylogenetic analysis (Phylogenetic signal), for example, using Xia’s test implemented in DAMBE (http://dambe.bio.uottawa.ca/dambe.asp). You need to test by potential effect of Saturation in DNA mutation before to perform the phylogenetic analyses.
Xia X, Xie Z, Salemi M, Chen L, Wang Y (2003) An index of substitution saturation and its application. Mol Phylogenet Evol 26: 1–7.
Xia X, Xie Z (2001) DAMBE: data analysis in molecular biology and evolution. J Hered 92: 371–373.
Diversity is not the only cause of "noise" in data. Other sources can be recombination, where one part of a sequence has one history while other parts have another history. It is possible to have data in the very nice, linear, nonsaturated range of the Ts and Tv vs distance plot that has a very poor consistency index. I don't think there is any SINGLE answer to how to detect good data, but I'd like to hear about many different methods that people are using.
Please do not forget that the consistency index (CI) measures the amount of homoplasy in a cladogram, however this measure only makes sense in a Parsimony analysis framework. Remember that in phylogenetic analysis based on explicit models (ML and Bayes) all sites are important to calculate the likelihood given the model of molecular evolution, and then the CI do no sense at all in these methods.
AliGROOVE – visualization of heterogeneous sequence divergence within multiple sequence alignments and detection of inflated branch support
http://www.biomedcentral.com/1471-2105/15/294
- Badges
- Science method
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics
- Phylogenetic Analysis