CBA Conference, day 2

Continuing with the 2015 CBA conference on species delimitation.

Today's first speaker was Sasha Mikheyev, who presented genome sequencing data on hybridisation zones in social insects. Most of his talk focused on Africanised bees in America, where he was lucky enough to find collaborators who had a time series of samples in the freezer capturing exactly the moment when the African genes swamped populations in the southern United States. Important research, but the results were still preliminary, so he could not go into what he ultimately wants to find out about the dynamics of hybridisation.

He ended his talk with a really weird story about an ant species in which queens are clones of their mothers, males are clones of their fathers (how does that work - is there a zygote but the female chromosomes are discarded?), and only the workers are half-half. This means that the males and females of the same "species" are genetically completely isolated, and indeed the male lineage is more closely related to a different species than to the females they have sex with.

You just can't make anything up that is more bizarre than what reality holds in stock for us.

Ziheng Yang followed with a very technical talk on simultaneous estimation of species limits and species relationships using his program BPP. Despite all the models and formulas he was, in my eyes at least, easy to follow, and it was actually very nice to have the nuts and bolts of coalescent species tree models explained.

One thing that stood out is the difficulty that the coalescent model causes for Monte Carlo Markov Chain analysis. Among the usual assumptions of the species tree coalescent is that there is no gene flow between species lineages. That means that the gene lineages are not allowed to coalesce downstream in time from where the species lineages containing them coalesce. So if you are estimating the species tree through MCMC and you change its topology, the current estimate of gene trees usually immediately becomes forbidden. I did not understand the algorithmic solution but the problem should be obvious.

This was followed by a very applied talk by Matthew Fujita who uses the BPP tool in practice to test species delimitation in squamates (lizards and relatives). The talk was very well presented and informative, but it soon occurred to me that one of the current buzzwords in my area is "integrative taxonomy". It means nothing more than using several independent types of data together, for example morphology and genetics, which we obviously already did when I was an undergrad. But you wouldn't know it from the way it is presented as revolutionary these days.

I also couldn't help but think that if we botanists used the same approach as herpetologists we would easily have ten million species of plants on the planet instead of a few hundred thousand. I was told that in his talk yesterday, which unfortunately I couldn't attend, Kevin Thiele apparently pointed out that biological diversity is fractal, and that you can always find finer scale patterns if you just look closely enough. That observation would seem to apply here.

Don't get me wrong - nothing against acknowledging the genetic distinctness of local populations or describing variability in ploidy level, but are all of those really species?

Well, I am not really qualified to assess squamate taxonomy, but at a purely intuitive level this one species on every rock approach has a similar feel to me as those European botanists who describe the 2,500th "'"'"species"'"'" of Taraxacum or Rubus (no, one set of scare quotes would not be sufficient to communicate my opinion of them). If those apomictic dandelions were sexual, the differences used to delimit them would obviously quickly disappear; if the lizards were moved around a bit, would most of them not likewise quickly introgress? It always comes back to what we consider a species, no matter how much we would like to avoid that discussion.

After morning tea Huw Ogilvie presented three different ways of testing species delimitation in the BEAST software package. The first part of the talk was basically an introduction into BEAST as such; the second part I found hard to follow. My notes peter out with the reference to the new package DISSECT, which I wrote down so that I can look it up for myself. It sounds interesting, but I worry that like most Bayesian methods it will probably need an exorbitant computing time for any larger-than-proof-of-concept dataset and come with numerous priors that I wouldn't know how to set.

Gilles Guillot then talked about STRUCTURE-style clustering analyses using combined data of different types, e.g. genetic and geographical or morphological. Here I had hoped for more practical information, for example on the lines of "this is my new software, and that is the kind of data you need to use it", but it was rather heavy on the theory. Also: yellow and light green formulas on grey background? Not a good idea.

I then had to skip the technical presentations of new software and attend a different set of talks outside of the conference. I somewhat regret it in the case of the BEAST template SNAPP. At least of BPP we had already heard a lot, even on practical application.

I was back for the "lightning talks". Whereas yesterday they were mostly students at early stages of their projects they featured more data and results today. First was Lars Jermiin, who presented a new model for rate heterogeneity. Andrew Hugall talked about using genomic data to study brittle stars and basket stars. I had never before heard of the latter, but they are as interesting as beautiful, and the talk was overall pretty awesome.

Joanne Birch showed morphological SNP data from Australian grasses with difficult species limits, but the results are still preliminary, and she focused a lot on data quality control. Amy Slender presented even more preliminary data on South Australian birds. Ian Brennan used barcode data for species delimitation in geckos, and broke a lance for that kind of data by pointing out that colleagues in many countries still cannot afford to get genomic data, so that is what there is. Rebecca Laver presented more geckos, this time from the Kimberleys, and Mozes Blom introduced us to rock- and tree-inhabiting lizards showing apparently rapid adaptation to either habitat.

One thing that I found curious today is that three or four speakers called the audience "you systematists" at various points, or even said that "the taxonomists" are complaining about their results. This would make sense if the philosopher John Wilkins did it, because he really is not a systematist himself, but the people who said these things today are using genomic data to figure out whether, say, gecko populations should be the same species or not, how they are related, and thus ultimately what rank they should have in the system of life. And they have published or are intending to publish taxonomic descriptions of new species.

So if I may ask: who do they think they are kidding?