Post date: Aug 11, 2020 6:31:55 PM
BEAST
- Try a run with all species considered as one species.
- If it does not work, try a run with all species considered as one species plus one or two outgroups.
- Modifications in xml files: mutation rate u is equal to mutation rate v. theta = 4Nu for diploids, 6Nu for triploids (check that) - change value accordingly.
- What is the unit of u? Is it per base mutation rate? This is very important, make sure it is correct.
- What is the part in the Prior window that says "snappprior convert to Nex".
- It is important to think about the prior on the mutation rate u: if we choose a log normal or a normal, we are either pulling in context order of magnitude versus values (lof versus not log). Normal is centered around the mean (the mean is the most liquely value) whereas uniform assumes equal probability around a range of different values around the mean. --> read the manual to understand how they are defining these things!
- Plot the different options.
- In the paper we can discuss that this is a very important choice and show the different results according to the different choices we make.
- Burning = do not sample in the first X chains because it is not relevant yet, it is too close to the prior values. You can assess when to start the burning by comparing different runs with the same parameters and look at when they converge.
- Run several chains at the same time, plus with less steps to see if it converges (100 000 steps).
- Modify wrapper to include emailing + paralelization )
ANGSD
- compare effective number of sites to contig size.
- Look: are they counting the sites where there is at least one read? or different?
- Is there an option to subset what is being counted as the number of effective sites?
CLUSTER
- copy my data in gompert group 2
- create symbolink link from my home directory