Could-do list - 11 March 2014
These are projects which mainly need writing.
- Main DISSECT paper. Submitted last week, waiting for reviews.
DISSECT on simulated data from Olave et al.
DISSECT looks very promising on this data, and this could make a (very?) short paper in Syst Biol.
Lots of waiting for MCMC runs, plus calculating evaluation metrics and making graphs.
I hope someone else does this.
Response to Mayrose et al.
I and others think this paper is flawed for various reasons and that a response should be written.
My strongest point is purely statistical: the way BISSE is used produces bias.
Second AlloppNET paper.
This is mostly written, and has been that way for nearly a year.
- DISSECT on Zeynep Aydin's data. Not sure what my contribution will be here, probably not much.
There are various ways in which the multispecies coalescent might be
modified or extended (other than for species delimitation).
PopsIO. That stands for 'population parameters integrated out,
and is a simpler multispecies coalescent model than *BEAST.
It might or might not mix better.
It is implemented in BEAST1, but hardly tested (it ran once).
Recombination. I think that the 'free recombination between genes' assumption
becomes increasingly dubious as the number of loci increases.
It should be possible to account for this via 'semi-linked' locus trees,
hopefully just by modifying the prior.
Need to read up on methods for detecting recombination within loci.
Gradual speciation. A soft barrier to gene flow might be good, the difficult
part is modelling cases where a partially split population splits further.
Badly behaved genes. If some genes are subject to late introgression, or
duplication/loss, or HGT, it would be nice to identify them and remove them
from the analysis in a principled Bayesian way.
Each gene could have a small probability of being 'bad', and if it was too
poor at fitting in with the rest, it would be removed.
Need MCMC moves for removing and adding genes from the 'good' set.
- Incorporate taxonomists' priors.
- More extensive simulations would be good, but I want others to do this.
Write a tutorial on priors for Bayesian phylogenetics. Somebody should do it.
Laplacian approximation instead of MCMC. Hard, lots of work.
Developmental plasticity and evolution.
Some simulations along the lines of Frank (2011) could be done,
but the most interesting aspects are very hard.
Mayrose et al, 2011. Recently Formed Polyploid Plants
Diversify at Lower Rates, Science, 333, 1257
S A Frank, 2011. Natural selection. II. Developmental variability and evolutionary rate.
Journal of evolutionary biology 24 (11), 2310-2320
Mary Jane West-Eberhard, 2003 Developmental plasticity and evolution, Oxford University
Press, New York.