Moving (carefully!) beyond molecular typing

When they take a look at teaching science and how kids learn, they will be much better prepared and much more focused.

While doing my PhD, I am a teacher assistant for undergrads students. When mining recent paper to help them along with their phylogenetic studies, I found one focused for veterinary students. When reading at the title (Applying phylogenetics analysis to viral livestock diseases: Moving beyond molecular typing) I found myself a few years ago starting going beyond dengue molecular typing and trying to get more worthwhile information from sequences.

From virology symposiums I remember guys presenting their main findings and adding at the end: We also sequenced a partial gene and build this tree topology with this blackbox program. What did they get anyway? Dunno, but they found it pretty cool. I also got familiar with: I want to do infer phylogeny; I get this from genbank, now where I should click? (sigh)

I checked this paper to ensure if it might be a useful resource for my undergrads; It is an overview of phylogenetic analysis in an attempt to improve incoming studies.

I kinda like paper writing style. Authors highlighted key stuff such as out-group selection being a critical issue (a very common undervalued step) and that gene tree is not necessarily representative of the species trees. Paragraphs about selection of the phylogenetic marker (we are poor, sequencing partial genes should work), constructing the data sets and selection of substitution model (Long live to Modeltest!) are also pretty nice. Authors suggested to be carefully when aligning (Keep going!).

Then, the paper titles Phylogenetic tree estimation and came my first concern:

…there are four main groups of statistical methods that can be used for reconstructing phylogenetic trees from molecular data, namely parsimony methods, distance methods, likelihood methods and Bayesian inference…

Is statistical placed here just to show us that statistic sounds cool anywhere or indeed are there nor statistical methods that can be used for reconstructing phylogenetic trees from molecular data (I want to know, really!). Anyhow they don’t use the common maximum parsimony term! (I always wanted to optimize data by average parsimony o minimal one). Distance methods weakness is stated (thanks god!) but not carefully suggestion are make about using Bayesian inference (we are all doomed! Priors are already here!); On the other hand they realize that every method has some strength and some weaknesses (with ML no one method being superior).

My main concern was about “evaluating the estimation”: Yup, lets place numbers in grouping, I like 100s (all we do)! You see, is real! (Statistics are like a drunk with a lamppost: used more for support than illumination!). The so called statistical method to asses support for grouping is recommended here, suggesting as “this surely is something” 90% or higher values, and 70% as “well, lower will be worst”. But take it easy, because you are still prone to a heart attack. It is intuitively suggested that Consensus tree is the new in the block approach to asses support in parsimony! R.I.P Bremer*! And what about Bayes, I did not seen suggestions here!

The remaining titles of the paper focus in phylogenetioc applications to livestock diseases studies. Which are not my concerns yet. Nonetheless, I consider this paper to be a good starting point to understand basics and further clarifications should go along with each title (Duh!).

* Bremer support is named after the Swedish botanist Kore Bremer, who devised the method, but it is also known as the “Decay Index”, for reasons that will become clear. The method asks the question: how much longer should the tree/cladogram be before a particular node collapses? The larger the number the stronger the support for that node.

Source: Olvera, A., & et al. (2010). Applying phylogenetic analysis to viral livestock diseases: Moving beyond molecular typing The Veterinary Journal, 184 (2), 130-137 DOI: 10.1016/j.tvjl.2009.02.015